Merge branch 'sched/core' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip...
authorMartin Schwidefsky <schwidefsky@de.ibm.com>
Mon, 19 Dec 2011 18:23:15 +0000 (19:23 +0100)
committerMartin Schwidefsky <schwidefsky@de.ibm.com>
Mon, 19 Dec 2011 18:23:15 +0000 (19:23 +0100)
Conflicts:
drivers/cpufreq/cpufreq_conservative.c
drivers/cpufreq/cpufreq_ondemand.c
drivers/macintosh/rack-meter.c
fs/proc/stat.c
fs/proc/uptime.c
kernel/sched/core.c

79 files changed:
Documentation/ABI/testing/sysfs-bus-rbd
arch/arm/Kconfig
arch/arm/include/asm/unwind.h
arch/arm/kernel/perf_event.c
arch/arm/kernel/setup.c
arch/arm/kernel/unwind.c
arch/ia64/include/asm/cputime.h
arch/powerpc/include/asm/cputime.h
arch/s390/include/asm/cputime.h
arch/x86/include/asm/e820.h
arch/x86/include/asm/efi.h
arch/x86/kernel/e820.c
arch/x86/kernel/setup.c
arch/x86/platform/efi/efi.c
arch/x86/platform/efi/efi_32.c
arch/x86/platform/efi/efi_64.c
drivers/block/rbd.c
drivers/cpufreq/cpufreq_conservative.c
drivers/cpufreq/cpufreq_ondemand.c
drivers/cpufreq/cpufreq_stats.c
drivers/hwmon/jz4740-hwmon.c
drivers/macintosh/rack-meter.c
drivers/mmc/card/block.c
drivers/mmc/core/core.c
drivers/mmc/core/mmc.c
drivers/mmc/host/mxcmmc.c
drivers/mmc/host/omap_hsmmc.c
drivers/mmc/host/sdhci-cns3xxx.c
drivers/mmc/host/sdhci-s3c.c
drivers/mmc/host/sh_mmcif.c
drivers/mmc/host/tmio_mmc_pio.c
fs/ceph/addr.c
fs/ceph/caps.c
fs/ceph/dir.c
fs/ceph/file.c
fs/ceph/inode.c
fs/ceph/ioctl.c
fs/ceph/mds_client.c
fs/ceph/mds_client.h
fs/ceph/snap.c
fs/ceph/super.c
fs/ceph/super.h
fs/ceph/xattr.c
fs/configfs/inode.c
fs/configfs/mount.c
fs/ext4/extents.c
fs/ext4/inode.c
fs/ext4/page-io.c
fs/ext4/super.c
fs/fs-writeback.c
fs/fuse/dev.c
fs/fuse/file.c
fs/fuse/inode.c
fs/ncpfs/inode.c
fs/proc/array.c
fs/proc/root.c
fs/proc/uptime.c
fs/ubifs/super.c
include/asm-generic/cputime.h
include/drm/drm_pciids.h
include/linux/log2.h
include/linux/mmc/card.h
include/linux/sched.h
ipc/mqueue.c
ipc/msgutil.c
kernel/acct.c
kernel/cpu.c
kernel/exit.c
kernel/fork.c
kernel/itimer.c
kernel/posix-cpu-timers.c
kernel/sched/core.c
kernel/sched/stats.h
kernel/signal.c
kernel/sys.c
kernel/tsacct.c
mm/filemap.c
mm/page-writeback.c
net/ceph/crush/mapper.c

index fa72ccb2282e77c879c0a7a135f6d86828a143cc..dbedafb095e24d3d3a8e2d93b6cbd727268d1754 100644 (file)
@@ -57,13 +57,6 @@ create_snap
 
         $ echo <snap-name> > /sys/bus/rbd/devices/<dev-id>/snap_create
 
-rollback_snap
-
-       Rolls back data to the specified snapshot. This goes over the entire
-       list of rados blocks and sends a rollback command to each.
-
-        $ echo <snap-name> > /sys/bus/rbd/devices/<dev-id>/snap_rollback
-
 snap_*
 
        A directory per each snapshot
index e084b7e981e8ff301aaae64ef3a2dc7e66701247..776d76b8cb695ff052f310d9f50a737209e96f1b 100644 (file)
@@ -220,8 +220,9 @@ config NEED_MACH_MEMORY_H
          be avoided when possible.
 
 config PHYS_OFFSET
-       hex "Physical address of main memory"
+       hex "Physical address of main memory" if MMU
        depends on !ARM_PATCH_PHYS_VIRT && !NEED_MACH_MEMORY_H
+       default DRAM_BASE if !MMU
        help
          Please provide the physical address corresponding to the
          location of main memory in your system.
index a5edf421005cce0d043cbb75be394253127a2582..d1c3f3a71c9454dd665b48a0cfc51bc954576a05 100644 (file)
@@ -30,14 +30,15 @@ enum unwind_reason_code {
 };
 
 struct unwind_idx {
-       unsigned long addr;
+       unsigned long addr_offset;
        unsigned long insn;
 };
 
 struct unwind_table {
        struct list_head list;
-       struct unwind_idx *start;
-       struct unwind_idx *stop;
+       const struct unwind_idx *start;
+       const struct unwind_idx *origin;
+       const struct unwind_idx *stop;
        unsigned long begin_addr;
        unsigned long end_addr;
 };
@@ -49,15 +50,6 @@ extern struct unwind_table *unwind_table_add(unsigned long start,
 extern void unwind_table_del(struct unwind_table *tab);
 extern void unwind_backtrace(struct pt_regs *regs, struct task_struct *tsk);
 
-#ifdef CONFIG_ARM_UNWIND
-extern int __init unwind_init(void);
-#else
-static inline int __init unwind_init(void)
-{
-       return 0;
-}
-#endif
-
 #endif /* !__ASSEMBLY__ */
 
 #ifdef CONFIG_ARM_UNWIND
index 8e9c98edc0682a8aa23790737a3b63b08dea847f..88b0941ce51ec674bbf46d604fcbb46f5dc25255 100644 (file)
@@ -640,6 +640,9 @@ static struct platform_device_id armpmu_plat_device_ids[] = {
 
 static int __devinit armpmu_device_probe(struct platform_device *pdev)
 {
+       if (!cpu_pmu)
+               return -ENODEV;
+
        cpu_pmu->plat_device = pdev;
        return 0;
 }
index 3448a3f9cc8c90ae71809075f1cc9c2d313fe167..8fc2c8fcbdc646a4a8babecbf4a758f88e6d8d60 100644 (file)
@@ -895,8 +895,6 @@ void __init setup_arch(char **cmdline_p)
 {
        struct machine_desc *mdesc;
 
-       unwind_init();
-
        setup_processor();
        mdesc = setup_machine_fdt(__atags_pointer);
        if (!mdesc)
@@ -904,6 +902,12 @@ void __init setup_arch(char **cmdline_p)
        machine_desc = mdesc;
        machine_name = mdesc->name;
 
+#ifdef CONFIG_ZONE_DMA
+       if (mdesc->dma_zone_size) {
+               extern unsigned long arm_dma_zone_size;
+               arm_dma_zone_size = mdesc->dma_zone_size;
+       }
+#endif
        if (mdesc->soft_reboot)
                reboot_setup("s");
 
@@ -934,12 +938,6 @@ void __init setup_arch(char **cmdline_p)
 
        tcm_init();
 
-#ifdef CONFIG_ZONE_DMA
-       if (mdesc->dma_zone_size) {
-               extern unsigned long arm_dma_zone_size;
-               arm_dma_zone_size = mdesc->dma_zone_size;
-       }
-#endif
 #ifdef CONFIG_MULTI_IRQ_HANDLER
        handle_arch_irq = mdesc->handle_irq;
 #endif
index e7e8365795c3d3272a4ef56d34d41662acd608b8..3f03fe0c3269d9993ece6e05ae47439605a4ff1c 100644 (file)
@@ -67,7 +67,7 @@ EXPORT_SYMBOL(__aeabi_unwind_cpp_pr2);
 
 struct unwind_ctrl_block {
        unsigned long vrs[16];          /* virtual register set */
-       unsigned long *insn;            /* pointer to the current instructions word */
+       const unsigned long *insn;      /* pointer to the current instructions word */
        int entries;                    /* number of entries left to interpret */
        int byte;                       /* current byte number in the instructions word */
 };
@@ -83,8 +83,9 @@ enum regs {
        PC = 15
 };
 
-extern struct unwind_idx __start_unwind_idx[];
-extern struct unwind_idx __stop_unwind_idx[];
+extern const struct unwind_idx __start_unwind_idx[];
+static const struct unwind_idx *__origin_unwind_idx;
+extern const struct unwind_idx __stop_unwind_idx[];
 
 static DEFINE_SPINLOCK(unwind_lock);
 static LIST_HEAD(unwind_tables);
@@ -98,45 +99,99 @@ static LIST_HEAD(unwind_tables);
 })
 
 /*
- * Binary search in the unwind index. The entries entries are
+ * Binary search in the unwind index. The entries are
  * guaranteed to be sorted in ascending order by the linker.
+ *
+ * start = first entry
+ * origin = first entry with positive offset (or stop if there is no such entry)
+ * stop - 1 = last entry
  */
-static struct unwind_idx *search_index(unsigned long addr,
-                                      struct unwind_idx *first,
-                                      struct unwind_idx *last)
+static const struct unwind_idx *search_index(unsigned long addr,
+                                      const struct unwind_idx *start,
+                                      const struct unwind_idx *origin,
+                                      const struct unwind_idx *stop)
 {
-       pr_debug("%s(%08lx, %p, %p)\n", __func__, addr, first, last);
+       unsigned long addr_prel31;
+
+       pr_debug("%s(%08lx, %p, %p, %p)\n",
+                       __func__, addr, start, origin, stop);
+
+       /*
+        * only search in the section with the matching sign. This way the
+        * prel31 numbers can be compared as unsigned longs.
+        */
+       if (addr < (unsigned long)start)
+               /* negative offsets: [start; origin) */
+               stop = origin;
+       else
+               /* positive offsets: [origin; stop) */
+               start = origin;
+
+       /* prel31 for address relavive to start */
+       addr_prel31 = (addr - (unsigned long)start) & 0x7fffffff;
 
-       if (addr < first->addr) {
+       while (start < stop - 1) {
+               const struct unwind_idx *mid = start + ((stop - start) >> 1);
+
+               /*
+                * As addr_prel31 is relative to start an offset is needed to
+                * make it relative to mid.
+                */
+               if (addr_prel31 - ((unsigned long)mid - (unsigned long)start) <
+                               mid->addr_offset)
+                       stop = mid;
+               else {
+                       /* keep addr_prel31 relative to start */
+                       addr_prel31 -= ((unsigned long)mid -
+                                       (unsigned long)start);
+                       start = mid;
+               }
+       }
+
+       if (likely(start->addr_offset <= addr_prel31))
+               return start;
+       else {
                pr_warning("unwind: Unknown symbol address %08lx\n", addr);
                return NULL;
-       } else if (addr >= last->addr)
-               return last;
+       }
+}
 
-       while (first < last - 1) {
-               struct unwind_idx *mid = first + ((last - first + 1) >> 1);
+static const struct unwind_idx *unwind_find_origin(
+               const struct unwind_idx *start, const struct unwind_idx *stop)
+{
+       pr_debug("%s(%p, %p)\n", __func__, start, stop);
+       while (start < stop - 1) {
+               const struct unwind_idx *mid = start + ((stop - start) >> 1);
 
-               if (addr < mid->addr)
-                       last = mid;
+               if (mid->addr_offset >= 0x40000000)
+                       /* negative offset */
+                       start = mid;
                else
-                       first = mid;
+                       /* positive offset */
+                       stop = mid;
        }
-
-       return first;
+       pr_debug("%s -> %p\n", __func__, stop);
+       return stop;
 }
 
-static struct unwind_idx *unwind_find_idx(unsigned long addr)
+static const struct unwind_idx *unwind_find_idx(unsigned long addr)
 {
-       struct unwind_idx *idx = NULL;
+       const struct unwind_idx *idx = NULL;
        unsigned long flags;
 
        pr_debug("%s(%08lx)\n", __func__, addr);
 
-       if (core_kernel_text(addr))
+       if (core_kernel_text(addr)) {
+               if (unlikely(!__origin_unwind_idx))
+                       __origin_unwind_idx =
+                               unwind_find_origin(__start_unwind_idx,
+                                               __stop_unwind_idx);
+
                /* main unwind table */
                idx = search_index(addr, __start_unwind_idx,
-                                  __stop_unwind_idx - 1);
-       else {
+                                  __origin_unwind_idx,
+                                  __stop_unwind_idx);
+       } else {
                /* module unwind tables */
                struct unwind_table *table;
 
@@ -145,7 +200,8 @@ static struct unwind_idx *unwind_find_idx(unsigned long addr)
                        if (addr >= table->begin_addr &&
                            addr < table->end_addr) {
                                idx = search_index(addr, table->start,
-                                                  table->stop - 1);
+                                                  table->origin,
+                                                  table->stop);
                                /* Move-to-front to exploit common traces */
                                list_move(&table->list, &unwind_tables);
                                break;
@@ -274,7 +330,7 @@ static int unwind_exec_insn(struct unwind_ctrl_block *ctrl)
 int unwind_frame(struct stackframe *frame)
 {
        unsigned long high, low;
-       struct unwind_idx *idx;
+       const struct unwind_idx *idx;
        struct unwind_ctrl_block ctrl;
 
        /* only go to a higher address on the stack */
@@ -399,7 +455,6 @@ struct unwind_table *unwind_table_add(unsigned long start, unsigned long size,
                                      unsigned long text_size)
 {
        unsigned long flags;
-       struct unwind_idx *idx;
        struct unwind_table *tab = kmalloc(sizeof(*tab), GFP_KERNEL);
 
        pr_debug("%s(%08lx, %08lx, %08lx, %08lx)\n", __func__, start, size,
@@ -408,15 +463,12 @@ struct unwind_table *unwind_table_add(unsigned long start, unsigned long size,
        if (!tab)
                return tab;
 
-       tab->start = (struct unwind_idx *)start;
-       tab->stop = (struct unwind_idx *)(start + size);
+       tab->start = (const struct unwind_idx *)start;
+       tab->stop = (const struct unwind_idx *)(start + size);
+       tab->origin = unwind_find_origin(tab->start, tab->stop);
        tab->begin_addr = text_addr;
        tab->end_addr = text_addr + text_size;
 
-       /* Convert the symbol addresses to absolute values */
-       for (idx = tab->start; idx < tab->stop; idx++)
-               idx->addr = prel31_to_addr(&idx->addr);
-
        spin_lock_irqsave(&unwind_lock, flags);
        list_add_tail(&tab->list, &unwind_tables);
        spin_unlock_irqrestore(&unwind_lock, flags);
@@ -437,16 +489,3 @@ void unwind_table_del(struct unwind_table *tab)
 
        kfree(tab);
 }
-
-int __init unwind_init(void)
-{
-       struct unwind_idx *idx;
-
-       /* Convert the symbol addresses to absolute values */
-       for (idx = __start_unwind_idx; idx < __stop_unwind_idx; idx++)
-               idx->addr = prel31_to_addr(&idx->addr);
-
-       pr_debug("unwind: ARM stack unwinding initialised\n");
-
-       return 0;
-}
index 6073b187528a26a8ac16826aa98cd7986c5724ed..461e52f0277fe494a245fda204d8e39cdf3ceb0e 100644 (file)
 #include <linux/jiffies.h>
 #include <asm/processor.h>
 
-typedef u64 cputime_t;
-typedef u64 cputime64_t;
+typedef u64 __nocast cputime_t;
+typedef u64 __nocast cputime64_t;
 
-#define cputime_zero                   ((cputime_t)0)
 #define cputime_one_jiffy              jiffies_to_cputime(1)
-#define cputime_max                    ((~((cputime_t)0) >> 1) - 1)
-#define cputime_add(__a, __b)          ((__a) +  (__b))
-#define cputime_sub(__a, __b)          ((__a) -  (__b))
-#define cputime_div(__a, __n)          ((__a) /  (__n))
-#define cputime_halve(__a)             ((__a) >> 1)
-#define cputime_eq(__a, __b)           ((__a) == (__b))
-#define cputime_gt(__a, __b)           ((__a) >  (__b))
-#define cputime_ge(__a, __b)           ((__a) >= (__b))
-#define cputime_lt(__a, __b)           ((__a) <  (__b))
-#define cputime_le(__a, __b)           ((__a) <= (__b))
-
-#define cputime64_zero                 ((cputime64_t)0)
-#define cputime64_add(__a, __b)                ((__a) + (__b))
-#define cputime64_sub(__a, __b)                ((__a) - (__b))
-#define cputime_to_cputime64(__ct)     (__ct)
 
 /*
  * Convert cputime <-> jiffies (HZ)
  */
-#define cputime_to_jiffies(__ct)       ((__ct) / (NSEC_PER_SEC / HZ))
-#define jiffies_to_cputime(__jif)      ((__jif) * (NSEC_PER_SEC / HZ))
-#define cputime64_to_jiffies64(__ct)   ((__ct) / (NSEC_PER_SEC / HZ))
-#define jiffies64_to_cputime64(__jif)  ((__jif) * (NSEC_PER_SEC / HZ))
+#define cputime_to_jiffies(__ct)       \
+       ((__force u64)(__ct) / (NSEC_PER_SEC / HZ))
+#define jiffies_to_cputime(__jif)      \
+       (__force cputime_t)((__jif) * (NSEC_PER_SEC / HZ))
+#define cputime64_to_jiffies64(__ct)   \
+       ((__force u64)(__ct) / (NSEC_PER_SEC / HZ))
+#define jiffies64_to_cputime64(__jif)  \
+       (__force cputime64_t)((__jif) * (NSEC_PER_SEC / HZ))
 
 /*
  * Convert cputime <-> microseconds
  */
-#define cputime_to_usecs(__ct)         ((__ct) / NSEC_PER_USEC)
-#define usecs_to_cputime(__usecs)      ((__usecs) * NSEC_PER_USEC)
+#define cputime_to_usecs(__ct)         \
+       ((__force u64)(__ct) / NSEC_PER_USEC)
+#define usecs_to_cputime(__usecs)      \
+       (__force cputime_t)((__usecs) * NSEC_PER_USEC)
 
 /*
  * Convert cputime <-> seconds
  */
-#define cputime_to_secs(__ct)          ((__ct) / NSEC_PER_SEC)
-#define secs_to_cputime(__secs)                ((__secs) * NSEC_PER_SEC)
+#define cputime_to_secs(__ct)          \
+       ((__force u64)(__ct) / NSEC_PER_SEC)
+#define secs_to_cputime(__secs)                \
+       (__force cputime_t)((__secs) * NSEC_PER_SEC)
 
 /*
  * Convert cputime <-> timespec (nsec)
  */
 static inline cputime_t timespec_to_cputime(const struct timespec *val)
 {
-       cputime_t ret = val->tv_sec * NSEC_PER_SEC;
-       return (ret + val->tv_nsec);
+       u64 ret = val->tv_sec * NSEC_PER_SEC + val->tv_nsec;
+       return (__force cputime_t) ret;
 }
 static inline void cputime_to_timespec(const cputime_t ct, struct timespec *val)
 {
-       val->tv_sec  = ct / NSEC_PER_SEC;
-       val->tv_nsec = ct % NSEC_PER_SEC;
+       val->tv_sec  = (__force u64) ct / NSEC_PER_SEC;
+       val->tv_nsec = (__force u64) ct % NSEC_PER_SEC;
 }
 
 /*
@@ -86,25 +78,28 @@ static inline void cputime_to_timespec(const cputime_t ct, struct timespec *val)
  */
 static inline cputime_t timeval_to_cputime(struct timeval *val)
 {
-       cputime_t ret = val->tv_sec * NSEC_PER_SEC;
-       return (ret + val->tv_usec * NSEC_PER_USEC);
+       u64 ret = val->tv_sec * NSEC_PER_SEC + val->tv_usec * NSEC_PER_USEC;
+       return (__force cputime_t) ret;
 }
 static inline void cputime_to_timeval(const cputime_t ct, struct timeval *val)
 {
-       val->tv_sec = ct / NSEC_PER_SEC;
-       val->tv_usec = (ct % NSEC_PER_SEC) / NSEC_PER_USEC;
+       val->tv_sec = (__force u64) ct / NSEC_PER_SEC;
+       val->tv_usec = ((__force u64) ct % NSEC_PER_SEC) / NSEC_PER_USEC;
 }
 
 /*
  * Convert cputime <-> clock (USER_HZ)
  */
-#define cputime_to_clock_t(__ct)       ((__ct) / (NSEC_PER_SEC / USER_HZ))
-#define clock_t_to_cputime(__x)                ((__x) * (NSEC_PER_SEC / USER_HZ))
+#define cputime_to_clock_t(__ct)       \
+       ((__force u64)(__ct) / (NSEC_PER_SEC / USER_HZ))
+#define clock_t_to_cputime(__x)                \
+       (__force cputime_t)((__x) * (NSEC_PER_SEC / USER_HZ))
 
 /*
  * Convert cputime64 to clock.
  */
-#define cputime64_to_clock_t(__ct)      cputime_to_clock_t((cputime_t)__ct)
+#define cputime64_to_clock_t(__ct)     \
+       cputime_to_clock_t((__force cputime_t)__ct)
 
 #endif /* CONFIG_VIRT_CPU_ACCOUNTING */
 #endif /* __IA64_CPUTIME_H */
index 1cf20bdfbecaada5bb2b88d439d776c32af2558c..e94935c520197d2d13af1eff0e8b149d4ad9b274 100644 (file)
@@ -29,25 +29,8 @@ static inline void setup_cputime_one_jiffy(void) { }
 #include <asm/time.h>
 #include <asm/param.h>
 
-typedef u64 cputime_t;
-typedef u64 cputime64_t;
-
-#define cputime_zero                   ((cputime_t)0)
-#define cputime_max                    ((~((cputime_t)0) >> 1) - 1)
-#define cputime_add(__a, __b)          ((__a) +  (__b))
-#define cputime_sub(__a, __b)          ((__a) -  (__b))
-#define cputime_div(__a, __n)          ((__a) /  (__n))
-#define cputime_halve(__a)             ((__a) >> 1)
-#define cputime_eq(__a, __b)           ((__a) == (__b))
-#define cputime_gt(__a, __b)           ((__a) >  (__b))
-#define cputime_ge(__a, __b)           ((__a) >= (__b))
-#define cputime_lt(__a, __b)           ((__a) <  (__b))
-#define cputime_le(__a, __b)           ((__a) <= (__b))
-
-#define cputime64_zero                 ((cputime64_t)0)
-#define cputime64_add(__a, __b)                ((__a) + (__b))
-#define cputime64_sub(__a, __b)                ((__a) - (__b))
-#define cputime_to_cputime64(__ct)     (__ct)
+typedef u64 __nocast cputime_t;
+typedef u64 __nocast cputime64_t;
 
 #ifdef __KERNEL__
 
@@ -65,7 +48,7 @@ DECLARE_PER_CPU(unsigned long, cputime_scaled_last_delta);
 
 static inline unsigned long cputime_to_jiffies(const cputime_t ct)
 {
-       return mulhdu(ct, __cputime_jiffies_factor);
+       return mulhdu((__force u64) ct, __cputime_jiffies_factor);
 }
 
 /* Estimate the scaled cputime by scaling the real cputime based on
@@ -74,14 +57,15 @@ static inline cputime_t cputime_to_scaled(const cputime_t ct)
 {
        if (cpu_has_feature(CPU_FTR_SPURR) &&
            __get_cpu_var(cputime_last_delta))
-               return ct * __get_cpu_var(cputime_scaled_last_delta) /
-                           __get_cpu_var(cputime_last_delta);
+               return (__force u64) ct *
+                       __get_cpu_var(cputime_scaled_last_delta) /
+                       __get_cpu_var(cputime_last_delta);
        return ct;
 }
 
 static inline cputime_t jiffies_to_cputime(const unsigned long jif)
 {
-       cputime_t ct;
+       u64 ct;
        unsigned long sec;
 
        /* have to be a little careful about overflow */
@@ -93,7 +77,7 @@ static inline cputime_t jiffies_to_cputime(const unsigned long jif)
        }
        if (sec)
                ct += (cputime_t) sec * tb_ticks_per_sec;
-       return ct;
+       return (__force cputime_t) ct;
 }
 
 static inline void setup_cputime_one_jiffy(void)
@@ -103,7 +87,7 @@ static inline void setup_cputime_one_jiffy(void)
 
 static inline cputime64_t jiffies64_to_cputime64(const u64 jif)
 {
-       cputime_t ct;
+       u64 ct;
        u64 sec;
 
        /* have to be a little careful about overflow */
@@ -114,13 +98,13 @@ static inline cputime64_t jiffies64_to_cputime64(const u64 jif)
                do_div(ct, HZ);
        }
        if (sec)
-               ct += (cputime_t) sec * tb_ticks_per_sec;
-       return ct;
+               ct += (u64) sec * tb_ticks_per_sec;
+       return (__force cputime64_t) ct;
 }
 
 static inline u64 cputime64_to_jiffies64(const cputime_t ct)
 {
-       return mulhdu(ct, __cputime_jiffies_factor);
+       return mulhdu((__force u64) ct, __cputime_jiffies_factor);
 }
 
 /*
@@ -130,12 +114,12 @@ extern u64 __cputime_msec_factor;
 
 static inline unsigned long cputime_to_usecs(const cputime_t ct)
 {
-       return mulhdu(ct, __cputime_msec_factor) * USEC_PER_MSEC;
+       return mulhdu((__force u64) ct, __cputime_msec_factor) * USEC_PER_MSEC;
 }
 
 static inline cputime_t usecs_to_cputime(const unsigned long us)
 {
-       cputime_t ct;
+       u64 ct;
        unsigned long sec;
 
        /* have to be a little careful about overflow */
@@ -147,7 +131,7 @@ static inline cputime_t usecs_to_cputime(const unsigned long us)
        }
        if (sec)
                ct += (cputime_t) sec * tb_ticks_per_sec;
-       return ct;
+       return (__force cputime_t) ct;
 }
 
 /*
@@ -157,12 +141,12 @@ extern u64 __cputime_sec_factor;
 
 static inline unsigned long cputime_to_secs(const cputime_t ct)
 {
-       return mulhdu(ct, __cputime_sec_factor);
+       return mulhdu((__force u64) ct, __cputime_sec_factor);
 }
 
 static inline cputime_t secs_to_cputime(const unsigned long sec)
 {
-       return (cputime_t) sec * tb_ticks_per_sec;
+       return (__force cputime_t)((u64) sec * tb_ticks_per_sec);
 }
 
 /*
@@ -170,7 +154,7 @@ static inline cputime_t secs_to_cputime(const unsigned long sec)
  */
 static inline void cputime_to_timespec(const cputime_t ct, struct timespec *p)
 {
-       u64 x = ct;
+       u64 x = (__force u64) ct;
        unsigned int frac;
 
        frac = do_div(x, tb_ticks_per_sec);
@@ -182,11 +166,11 @@ static inline void cputime_to_timespec(const cputime_t ct, struct timespec *p)
 
 static inline cputime_t timespec_to_cputime(const struct timespec *p)
 {
-       cputime_t ct;
+       u64 ct;
 
        ct = (u64) p->tv_nsec * tb_ticks_per_sec;
        do_div(ct, 1000000000);
-       return ct + (u64) p->tv_sec * tb_ticks_per_sec;
+       return (__force cputime_t)(ct + (u64) p->tv_sec * tb_ticks_per_sec);
 }
 
 /*
@@ -194,7 +178,7 @@ static inline cputime_t timespec_to_cputime(const struct timespec *p)
  */
 static inline void cputime_to_timeval(const cputime_t ct, struct timeval *p)
 {
-       u64 x = ct;
+       u64 x = (__force u64) ct;
        unsigned int frac;
 
        frac = do_div(x, tb_ticks_per_sec);
@@ -206,11 +190,11 @@ static inline void cputime_to_timeval(const cputime_t ct, struct timeval *p)
 
 static inline cputime_t timeval_to_cputime(const struct timeval *p)
 {
-       cputime_t ct;
+       u64 ct;
 
        ct = (u64) p->tv_usec * tb_ticks_per_sec;
        do_div(ct, 1000000);
-       return ct + (u64) p->tv_sec * tb_ticks_per_sec;
+       return (__force cputime_t)(ct + (u64) p->tv_sec * tb_ticks_per_sec);
 }
 
 /*
@@ -220,12 +204,12 @@ extern u64 __cputime_clockt_factor;
 
 static inline unsigned long cputime_to_clock_t(const cputime_t ct)
 {
-       return mulhdu(ct, __cputime_clockt_factor);
+       return mulhdu((__force u64) ct, __cputime_clockt_factor);
 }
 
 static inline cputime_t clock_t_to_cputime(const unsigned long clk)
 {
-       cputime_t ct;
+       u64 ct;
        unsigned long sec;
 
        /* have to be a little careful about overflow */
@@ -236,8 +220,8 @@ static inline cputime_t clock_t_to_cputime(const unsigned long clk)
                do_div(ct, USER_HZ);
        }
        if (sec)
-               ct += (cputime_t) sec * tb_ticks_per_sec;
-       return ct;
+               ct += (u64) sec * tb_ticks_per_sec;
+       return (__force cputime_t) ct;
 }
 
 #define cputime64_to_clock_t(ct)       cputime_to_clock_t((cputime_t)(ct))
index 0814348782966ac99c571990c89975d06aedffe1..0887a0463e338ce9ae50bbc9661ade13ac8aa75b 100644 (file)
 
 /* We want to use full resolution of the CPU timer: 2**-12 micro-seconds. */
 
-typedef unsigned long long cputime_t;
-typedef unsigned long long cputime64_t;
+typedef unsigned long long __nocast cputime_t;
+typedef unsigned long long __nocast cputime64_t;
 
-#ifndef __s390x__
-
-static inline unsigned int
-__div(unsigned long long n, unsigned int base)
+static inline unsigned long __div(unsigned long long n, unsigned long base)
 {
+#ifndef __s390x__
        register_pair rp;
 
        rp.pair = n >> 1;
        asm ("dr %0,%1" : "+d" (rp) : "d" (base >> 1));
        return rp.subreg.odd;
+#else /* __s390x__ */
+       return n / base;
+#endif /* __s390x__ */
 }
 
-#else /* __s390x__ */
+#define cputime_one_jiffy              jiffies_to_cputime(1)
 
-static inline unsigned int
-__div(unsigned long long n, unsigned int base)
+/*
+ * Convert cputime to jiffies and back.
+ */
+static inline unsigned long cputime_to_jiffies(const cputime_t cputime)
 {
-       return n / base;
+       return __div((__force unsigned long long) cputime, 4096000000ULL / HZ);
 }
 
-#endif /* __s390x__ */
+static inline cputime_t jiffies_to_cputime(const unsigned int jif)
+{
+       return (__force cputime_t)(jif * (4096000000ULL / HZ));
+}
 
-#define cputime_zero                   (0ULL)
-#define cputime_one_jiffy              jiffies_to_cputime(1)
-#define cputime_max                    ((~0UL >> 1) - 1)
-#define cputime_add(__a, __b)          ((__a) +  (__b))
-#define cputime_sub(__a, __b)          ((__a) -  (__b))
-#define cputime_div(__a, __n) ({               \
-       unsigned long long __div = (__a);       \
-       do_div(__div,__n);                      \
-       __div;                                  \
-})
-#define cputime_halve(__a)             ((__a) >> 1)
-#define cputime_eq(__a, __b)           ((__a) == (__b))
-#define cputime_gt(__a, __b)           ((__a) >  (__b))
-#define cputime_ge(__a, __b)           ((__a) >= (__b))
-#define cputime_lt(__a, __b)           ((__a) <  (__b))
-#define cputime_le(__a, __b)           ((__a) <= (__b))
-#define cputime_to_jiffies(__ct)       (__div((__ct), 4096000000ULL / HZ))
-#define cputime_to_scaled(__ct)                (__ct)
-#define jiffies_to_cputime(__hz)       ((cputime_t)(__hz) * (4096000000ULL / HZ))
-
-#define cputime64_zero                 (0ULL)
-#define cputime64_add(__a, __b)                ((__a) + (__b))
-#define cputime_to_cputime64(__ct)     (__ct)
-
-static inline u64
-cputime64_to_jiffies64(cputime64_t cputime)
-{
-       do_div(cputime, 4096000000ULL / HZ);
-       return cputime;
+static inline u64 cputime64_to_jiffies64(cputime64_t cputime)
+{
+       unsigned long long jif = (__force unsigned long long) cputime;
+       do_div(jif, 4096000000ULL / HZ);
+       return jif;
+}
+
+static inline cputime64_t jiffies64_to_cputime64(const u64 jif)
+{
+       return (__force cputime64_t)(jif * (4096000000ULL / HZ));
 }
 
 /*
  * Convert cputime to microseconds and back.
  */
-static inline unsigned int
-cputime_to_usecs(const cputime_t cputime)
+static inline unsigned int cputime_to_usecs(const cputime_t cputime)
 {
-       return cputime_div(cputime, 4096);
+       return (__force unsigned long long) cputime >> 12;
 }
 
-static inline cputime_t
-usecs_to_cputime(const unsigned int m)
+static inline cputime_t usecs_to_cputime(const unsigned int m)
 {
-       return (cputime_t) m * 4096;
+       return (__force cputime_t)(m * 4096ULL);
 }
 
 /*
  * Convert cputime to milliseconds and back.
  */
-static inline unsigned int
-cputime_to_secs(const cputime_t cputime)
+static inline unsigned int cputime_to_secs(const cputime_t cputime)
 {
-       return __div(cputime, 2048000000) >> 1;
+       return __div((__force unsigned long long) cputime, 2048000000) >> 1;
 }
 
-static inline cputime_t
-secs_to_cputime(const unsigned int s)
+static inline cputime_t secs_to_cputime(const unsigned int s)
 {
-       return (cputime_t) s * 4096000000ULL;
+       return (__force cputime_t)(s * 4096000000ULL);
 }
 
 /*
  * Convert cputime to timespec and back.
  */
-static inline cputime_t
-timespec_to_cputime(const struct timespec *value)
+static inline cputime_t timespec_to_cputime(const struct timespec *value)
 {
-       return value->tv_nsec * 4096 / 1000 + (u64) value->tv_sec * 4096000000ULL;
+       unsigned long long ret = value->tv_sec * 4096000000ULL;
+       return (__force cputime_t)(ret + value->tv_nsec * 4096 / 1000);
 }
 
-static inline void
-cputime_to_timespec(const cputime_t cputime, struct timespec *value)
+static inline void cputime_to_timespec(const cputime_t cputime,
+                                      struct timespec *value)
 {
+       unsigned long long __cputime = (__force unsigned long long) cputime;
 #ifndef __s390x__
        register_pair rp;
 
-       rp.pair = cputime >> 1;
+       rp.pair = __cputime >> 1;
        asm ("dr %0,%1" : "+d" (rp) : "d" (2048000000UL));
        value->tv_nsec = rp.subreg.even * 1000 / 4096;
        value->tv_sec = rp.subreg.odd;
 #else
-       value->tv_nsec = (cputime % 4096000000ULL) * 1000 / 4096;
-       value->tv_sec = cputime / 4096000000ULL;
+       value->tv_nsec = (__cputime % 4096000000ULL) * 1000 / 4096;
+       value->tv_sec = __cputime / 4096000000ULL;
 #endif
 }
 
@@ -132,50 +116,52 @@ cputime_to_timespec(const cputime_t cputime, struct timespec *value)
  * Since cputime and timeval have the same resolution (microseconds)
  * this is easy.
  */
-static inline cputime_t
-timeval_to_cputime(const struct timeval *value)
+static inline cputime_t timeval_to_cputime(const struct timeval *value)
 {
-       return value->tv_usec * 4096 + (u64) value->tv_sec * 4096000000ULL;
+       unsigned long long ret = value->tv_sec * 4096000000ULL;
+       return (__force cputime_t)(ret + value->tv_usec * 4096ULL);
 }
 
-static inline void
-cputime_to_timeval(const cputime_t cputime, struct timeval *value)
+static inline void cputime_to_timeval(const cputime_t cputime,
+                                     struct timeval *value)
 {
+       unsigned long long __cputime = (__force unsigned long long) cputime;
 #ifndef __s390x__
        register_pair rp;
 
-       rp.pair = cputime >> 1;
+       rp.pair = __cputime >> 1;
        asm ("dr %0,%1" : "+d" (rp) : "d" (2048000000UL));
        value->tv_usec = rp.subreg.even / 4096;
        value->tv_sec = rp.subreg.odd;
 #else
-       value->tv_usec = (cputime % 4096000000ULL) / 4096;
-       value->tv_sec = cputime / 4096000000ULL;
+       value->tv_usec = (__cputime % 4096000000ULL) / 4096;
+       value->tv_sec = __cputime / 4096000000ULL;
 #endif
 }
 
 /*
  * Convert cputime to clock and back.
  */
-static inline clock_t
-cputime_to_clock_t(cputime_t cputime)
+static inline clock_t cputime_to_clock_t(cputime_t cputime)
 {
-       return cputime_div(cputime, 4096000000ULL / USER_HZ);
+       unsigned long long clock = (__force unsigned long long) cputime;
+       do_div(clock, 4096000000ULL / USER_HZ);
+       return clock;
 }
 
-static inline cputime_t
-clock_t_to_cputime(unsigned long x)
+static inline cputime_t clock_t_to_cputime(unsigned long x)
 {
-       return (cputime_t) x * (4096000000ULL / USER_HZ);
+       return (__force cputime_t)(x * (4096000000ULL / USER_HZ));
 }
 
 /*
  * Convert cputime64 to clock.
  */
-static inline clock_t
-cputime64_to_clock_t(cputime64_t cputime)
+static inline clock_t cputime64_to_clock_t(cputime64_t cputime)
 {
-       return cputime_div(cputime, 4096000000ULL / USER_HZ);
+       unsigned long long clock = (__force unsigned long long) cputime;
+       do_div(clock, 4096000000ULL / USER_HZ);
+       return clock;
 }
 
 struct s390_idle_data {
index c9547033e38e8fee78964ffda671d9e3402a4455..908b96957d88adf11694f6e83652b47d06d3b8f0 100644 (file)
  */
 #define E820_RESERVED_KERN        128
 
-/*
- * Address ranges that need to be mapped by the kernel direct
- * mapping. This is used to make sure regions such as
- * EFI_RUNTIME_SERVICES_DATA are directly mapped. See setup_arch().
- */
-#define E820_RESERVED_EFI         129
-
 #ifndef __ASSEMBLY__
 #include <linux/types.h>
 struct e820entry {
@@ -122,7 +115,6 @@ static inline void early_memtest(unsigned long start, unsigned long end)
 }
 #endif
 
-extern unsigned long e820_end_pfn(unsigned long limit_pfn, unsigned type);
 extern unsigned long e820_end_of_ram_pfn(void);
 extern unsigned long e820_end_of_low_ram_pfn(void);
 extern u64 early_reserve_e820(u64 startt, u64 sizet, u64 align);
index b8d8bfcd44a95f48b15537b555636bbbb32c57b9..7093e4a6a0bc6dd5644b2b14cbef9ef5f4085eef 100644 (file)
@@ -33,6 +33,8 @@ extern unsigned long asmlinkage efi_call_phys(void *, ...);
 #define efi_call_virt6(f, a1, a2, a3, a4, a5, a6)      \
        efi_call_virt(f, a1, a2, a3, a4, a5, a6)
 
+#define efi_ioremap(addr, size, type)          ioremap_cache(addr, size)
+
 #else /* !CONFIG_X86_32 */
 
 extern u64 efi_call0(void *fp);
@@ -82,6 +84,9 @@ extern u64 efi_call6(void *fp, u64 arg1, u64 arg2, u64 arg3,
        efi_call6((void *)(efi.systab->runtime->f), (u64)(a1), (u64)(a2), \
                  (u64)(a3), (u64)(a4), (u64)(a5), (u64)(a6))
 
+extern void __iomem *efi_ioremap(unsigned long addr, unsigned long size,
+                                u32 type);
+
 #endif /* CONFIG_X86_32 */
 
 extern int add_efi_memmap;
index 65ffd110a81bc95491fb13ae7c87cd7be1738f93..303a0e48f076feb3feb522d4052ac4b958995d42 100644 (file)
@@ -135,7 +135,6 @@ static void __init e820_print_type(u32 type)
                printk(KERN_CONT "(usable)");
                break;
        case E820_RESERVED:
-       case E820_RESERVED_EFI:
                printk(KERN_CONT "(reserved)");
                break;
        case E820_ACPI:
@@ -784,7 +783,7 @@ u64 __init early_reserve_e820(u64 startt, u64 sizet, u64 align)
 /*
  * Find the highest page frame number we have available
  */
-unsigned long __init e820_end_pfn(unsigned long limit_pfn, unsigned type)
+static unsigned long __init e820_end_pfn(unsigned long limit_pfn, unsigned type)
 {
        int i;
        unsigned long last_pfn = 0;
index 9a9e40fb091ccbf0c79885dbbfd7d2e64b125d45..cf0ef986cb6dff51348c17c691491f6f48c61a60 100644 (file)
@@ -691,8 +691,6 @@ early_param("reservelow", parse_reservelow);
 
 void __init setup_arch(char **cmdline_p)
 {
-       unsigned long end_pfn;
-
 #ifdef CONFIG_X86_32
        memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
        visws_early_detect();
@@ -934,24 +932,7 @@ void __init setup_arch(char **cmdline_p)
        init_gbpages();
 
        /* max_pfn_mapped is updated here */
-       end_pfn = max_low_pfn;
-
-#ifdef CONFIG_X86_64
-       /*
-        * There may be regions after the last E820_RAM region that we
-        * want to include in the kernel direct mapping, such as
-        * EFI_RUNTIME_SERVICES_DATA.
-        */
-       if (efi_enabled) {
-               unsigned long efi_end;
-
-               efi_end = e820_end_pfn(MAXMEM>>PAGE_SHIFT, E820_RESERVED_EFI);
-               if (efi_end > max_low_pfn)
-                       end_pfn = efi_end;
-       }
-#endif
-
-       max_low_pfn_mapped = init_memory_mapping(0, end_pfn << PAGE_SHIFT);
+       max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
        max_pfn_mapped = max_low_pfn_mapped;
 
 #ifdef CONFIG_X86_64
index c9718a16be158bfa994e774b1b1ad2dd079c68cb..37718f0f053d53346566c80958885f159b61dcde 100644 (file)
@@ -323,13 +323,10 @@ static void __init do_add_efi_memmap(void)
                case EFI_UNUSABLE_MEMORY:
                        e820_type = E820_UNUSABLE;
                        break;
-               case EFI_RUNTIME_SERVICES_DATA:
-                       e820_type = E820_RESERVED_EFI;
-                       break;
                default:
                        /*
                         * EFI_RESERVED_TYPE EFI_RUNTIME_SERVICES_CODE
-                        * EFI_MEMORY_MAPPED_IO
+                        * EFI_RUNTIME_SERVICES_DATA EFI_MEMORY_MAPPED_IO
                         * EFI_MEMORY_MAPPED_IO_PORT_SPACE EFI_PAL_CODE
                         */
                        e820_type = E820_RESERVED;
@@ -674,21 +671,10 @@ void __init efi_enter_virtual_mode(void)
                end_pfn = PFN_UP(end);
                if (end_pfn <= max_low_pfn_mapped
                    || (end_pfn > (1UL << (32 - PAGE_SHIFT))
-                       && end_pfn <= max_pfn_mapped)) {
+                       && end_pfn <= max_pfn_mapped))
                        va = __va(md->phys_addr);
-
-                       if (!(md->attribute & EFI_MEMORY_WB)) {
-                               addr = (u64) (unsigned long)va;
-                               npages = md->num_pages;
-                               memrange_efi_to_native(&addr, &npages);
-                               set_memory_uc(addr, npages);
-                       }
-               } else {
-                       if (!(md->attribute & EFI_MEMORY_WB))
-                               va = ioremap_nocache(md->phys_addr, size);
-                       else
-                               va = ioremap_cache(md->phys_addr, size);
-               }
+               else
+                       va = efi_ioremap(md->phys_addr, size, md->type);
 
                md->virt_addr = (u64) (unsigned long) va;
 
@@ -698,6 +684,13 @@ void __init efi_enter_virtual_mode(void)
                        continue;
                }
 
+               if (!(md->attribute & EFI_MEMORY_WB)) {
+                       addr = md->virt_addr;
+                       npages = md->num_pages;
+                       memrange_efi_to_native(&addr, &npages);
+                       set_memory_uc(addr, npages);
+               }
+
                systab = (u64) (unsigned long) efi_phys.systab;
                if (md->phys_addr <= systab && systab < end) {
                        systab += md->virt_addr - md->phys_addr;
index e36bf714cb7727352c237d6bffa0d838d2c1c08e..40e446941dd7eceb587b2b4572c96279631e5f56 100644 (file)
  */
 
 static unsigned long efi_rt_eflags;
-static pgd_t efi_bak_pg_dir_pointer[2];
 
 void efi_call_phys_prelog(void)
 {
-       unsigned long cr4;
-       unsigned long temp;
        struct desc_ptr gdt_descr;
 
        local_irq_save(efi_rt_eflags);
 
-       /*
-        * If I don't have PAE, I should just duplicate two entries in page
-        * directory. If I have PAE, I just need to duplicate one entry in
-        * page directory.
-        */
-       cr4 = read_cr4_safe();
-
-       if (cr4 & X86_CR4_PAE) {
-               efi_bak_pg_dir_pointer[0].pgd =
-                   swapper_pg_dir[pgd_index(0)].pgd;
-               swapper_pg_dir[0].pgd =
-                   swapper_pg_dir[pgd_index(PAGE_OFFSET)].pgd;
-       } else {
-               efi_bak_pg_dir_pointer[0].pgd =
-                   swapper_pg_dir[pgd_index(0)].pgd;
-               efi_bak_pg_dir_pointer[1].pgd =
-                   swapper_pg_dir[pgd_index(0x400000)].pgd;
-               swapper_pg_dir[pgd_index(0)].pgd =
-                   swapper_pg_dir[pgd_index(PAGE_OFFSET)].pgd;
-               temp = PAGE_OFFSET + 0x400000;
-               swapper_pg_dir[pgd_index(0x400000)].pgd =
-                   swapper_pg_dir[pgd_index(temp)].pgd;
-       }
-
-       /*
-        * After the lock is released, the original page table is restored.
-        */
+       load_cr3(initial_page_table);
        __flush_tlb_all();
 
        gdt_descr.address = __pa(get_cpu_gdt_table(0));
@@ -85,28 +56,13 @@ void efi_call_phys_prelog(void)
 
 void efi_call_phys_epilog(void)
 {
-       unsigned long cr4;
        struct desc_ptr gdt_descr;
 
        gdt_descr.address = (unsigned long)get_cpu_gdt_table(0);
        gdt_descr.size = GDT_SIZE - 1;
        load_gdt(&gdt_descr);
 
-       cr4 = read_cr4_safe();
-
-       if (cr4 & X86_CR4_PAE) {
-               swapper_pg_dir[pgd_index(0)].pgd =
-                   efi_bak_pg_dir_pointer[0].pgd;
-       } else {
-               swapper_pg_dir[pgd_index(0)].pgd =
-                   efi_bak_pg_dir_pointer[0].pgd;
-               swapper_pg_dir[pgd_index(0x400000)].pgd =
-                   efi_bak_pg_dir_pointer[1].pgd;
-       }
-
-       /*
-        * After the lock is released, the original page table is restored.
-        */
+       load_cr3(swapper_pg_dir);
        __flush_tlb_all();
 
        local_irq_restore(efi_rt_eflags);
index 312250c6b2de3078bc47562c6b3034e53e74e4f3..ac3aa54e26546ba5cb4121eba0c58ca00f06ea82 100644 (file)
@@ -80,3 +80,20 @@ void __init efi_call_phys_epilog(void)
        local_irq_restore(efi_flags);
        early_code_mapping_set_exec(0);
 }
+
+void __iomem *__init efi_ioremap(unsigned long phys_addr, unsigned long size,
+                                u32 type)
+{
+       unsigned long last_map_pfn;
+
+       if (type == EFI_MEMORY_MAPPED_IO)
+               return ioremap(phys_addr, size);
+
+       last_map_pfn = init_memory_mapping(phys_addr, phys_addr + size);
+       if ((last_map_pfn << PAGE_SHIFT) < phys_addr + size) {
+               unsigned long top = last_map_pfn << PAGE_SHIFT;
+               efi_ioremap(top, size - (top - phys_addr), type);
+       }
+
+       return (void __iomem *)__va(phys_addr);
+}
index 65cc424359b05e735703b1b9a74277cb1f1359a9..148ab944378d57bdaec596e96bed6dc131777a28 100644 (file)
@@ -183,10 +183,6 @@ static LIST_HEAD(rbd_client_list);      /* clients */
 
 static int __rbd_init_snaps_header(struct rbd_device *rbd_dev);
 static void rbd_dev_release(struct device *dev);
-static ssize_t rbd_snap_rollback(struct device *dev,
-                                struct device_attribute *attr,
-                                const char *buf,
-                                size_t size);
 static ssize_t rbd_snap_add(struct device *dev,
                            struct device_attribute *attr,
                            const char *buf,
@@ -461,6 +457,10 @@ static int rbd_header_from_disk(struct rbd_image_header *header,
        u32 snap_count = le32_to_cpu(ondisk->snap_count);
        int ret = -ENOMEM;
 
+       if (memcmp(ondisk, RBD_HEADER_TEXT, sizeof(RBD_HEADER_TEXT))) {
+               return -ENXIO;
+       }
+
        init_rwsem(&header->snap_rwsem);
        header->snap_names_len = le64_to_cpu(ondisk->snap_names_len);
        header->snapc = kmalloc(sizeof(struct ceph_snap_context) +
@@ -1355,32 +1355,6 @@ fail:
        return ret;
 }
 
-/*
- * Request sync osd rollback
- */
-static int rbd_req_sync_rollback_obj(struct rbd_device *dev,
-                                    u64 snapid,
-                                    const char *obj)
-{
-       struct ceph_osd_req_op *ops;
-       int ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_ROLLBACK, 0);
-       if (ret < 0)
-               return ret;
-
-       ops[0].snap.snapid = snapid;
-
-       ret = rbd_req_sync_op(dev, NULL,
-                              CEPH_NOSNAP,
-                              0,
-                              CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK,
-                              ops,
-                              1, obj, 0, 0, NULL, NULL, NULL);
-
-       rbd_destroy_ops(ops);
-
-       return ret;
-}
-
 /*
  * Request sync osd read
  */
@@ -1610,8 +1584,13 @@ static int rbd_read_header(struct rbd_device *rbd_dev,
                        goto out_dh;
 
                rc = rbd_header_from_disk(header, dh, snap_count, GFP_KERNEL);
-               if (rc < 0)
+               if (rc < 0) {
+                       if (rc == -ENXIO) {
+                               pr_warning("unrecognized header format"
+                                          " for image %s", rbd_dev->obj);
+                       }
                        goto out_dh;
+               }
 
                if (snap_count != header->total_snaps) {
                        snap_count = header->total_snaps;
@@ -1882,7 +1861,6 @@ static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL);
 static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh);
 static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL);
 static DEVICE_ATTR(create_snap, S_IWUSR, NULL, rbd_snap_add);
-static DEVICE_ATTR(rollback_snap, S_IWUSR, NULL, rbd_snap_rollback);
 
 static struct attribute *rbd_attrs[] = {
        &dev_attr_size.attr,
@@ -1893,7 +1871,6 @@ static struct attribute *rbd_attrs[] = {
        &dev_attr_current_snap.attr,
        &dev_attr_refresh.attr,
        &dev_attr_create_snap.attr,
-       &dev_attr_rollback_snap.attr,
        NULL
 };
 
@@ -2424,64 +2401,6 @@ err_unlock:
        return ret;
 }
 
-static ssize_t rbd_snap_rollback(struct device *dev,
-                                struct device_attribute *attr,
-                                const char *buf,
-                                size_t count)
-{
-       struct rbd_device *rbd_dev = dev_to_rbd(dev);
-       int ret;
-       u64 snapid;
-       u64 cur_ofs;
-       char *seg_name = NULL;
-       char *snap_name = kmalloc(count + 1, GFP_KERNEL);
-       ret = -ENOMEM;
-       if (!snap_name)
-               return ret;
-
-       /* parse snaps add command */
-       snprintf(snap_name, count, "%s", buf);
-       seg_name = kmalloc(RBD_MAX_SEG_NAME_LEN + 1, GFP_NOIO);
-       if (!seg_name)
-               goto done;
-
-       mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING);
-
-       ret = snap_by_name(&rbd_dev->header, snap_name, &snapid, NULL);
-       if (ret < 0)
-               goto done_unlock;
-
-       dout("snapid=%lld\n", snapid);
-
-       cur_ofs = 0;
-       while (cur_ofs < rbd_dev->header.image_size) {
-               cur_ofs += rbd_get_segment(&rbd_dev->header,
-                                          rbd_dev->obj,
-                                          cur_ofs, (u64)-1,
-                                          seg_name, NULL);
-               dout("seg_name=%s\n", seg_name);
-
-               ret = rbd_req_sync_rollback_obj(rbd_dev, snapid, seg_name);
-               if (ret < 0)
-                       pr_warning("could not roll back obj %s err=%d\n",
-                                  seg_name, ret);
-       }
-
-       ret = __rbd_update_snaps(rbd_dev);
-       if (ret < 0)
-               goto done_unlock;
-
-       ret = count;
-
-done_unlock:
-       mutex_unlock(&ctl_mutex);
-done:
-       kfree(seg_name);
-       kfree(snap_name);
-
-       return ret;
-}
-
 static struct bus_attribute rbd_bus_attrs[] = {
        __ATTR(add, S_IWUSR, NULL, rbd_add),
        __ATTR(remove, S_IWUSR, NULL, rbd_remove),
index 118bff73fed36786984aa34ba33d0b5eec131af4..235a340e81f20bcb63805f9c911f4fd99eafeab8 100644 (file)
@@ -98,19 +98,19 @@ static struct dbs_tuners {
 static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
 {
        u64 idle_time;
-       cputime64_t cur_wall_time;
+       u64 cur_wall_time;
        u64 busy_time;
 
        cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
-       busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER] +
-                   kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
 
+       busy_time  = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
+       busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
        busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
        busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
        busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
        busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
 
-       idle_time = cputime64_sub(cur_wall_time, busy_time);
+       idle_time = cur_wall_time - busy_time;
        if (wall)
                *wall = jiffies_to_usecs(cur_wall_time);
 
@@ -352,12 +352,12 @@ static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info)
 
                cur_idle_time = get_cpu_idle_time(j, &cur_wall_time);
 
-               wall_time = (unsigned int) cputime64_sub(cur_wall_time,
-                               j_dbs_info->prev_cpu_wall);
+               wall_time = (unsigned int)
+                       (cur_wall_time - j_dbs_info->prev_cpu_wall);
                j_dbs_info->prev_cpu_wall = cur_wall_time;
 
-               idle_time = (unsigned int) cputime64_sub(cur_idle_time,
-                               j_dbs_info->prev_cpu_idle);
+               idle_time = (unsigned int)
+                       (cur_idle_time - j_dbs_info->prev_cpu_idle);
                j_dbs_info->prev_cpu_idle = cur_idle_time;
 
                if (dbs_tuners_ins.ignore_nice) {
index f3d327cee43f891b0c49614409266cc944947f49..3d679eee70a137266df009c1d91ae20535409a33 100644 (file)
@@ -122,19 +122,19 @@ static struct dbs_tuners {
 static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
 {
        u64 idle_time;
-       cputime64_t cur_wall_time;
+       u64 cur_wall_time;
        u64 busy_time;
 
        cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
-       busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER] +
-                   kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
 
+       busy_time  = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
+       busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
        busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
        busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
        busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
        busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
 
-       idle_time = cputime64_sub(cur_wall_time, busy_time);
+       idle_time = cur_wall_time - busy_time;
        if (wall)
                *wall = jiffies_to_usecs(cur_wall_time);
 
@@ -441,16 +441,16 @@ static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info)
                cur_idle_time = get_cpu_idle_time(j, &cur_wall_time);
                cur_iowait_time = get_cpu_iowait_time(j, &cur_wall_time);
 
-               wall_time = (unsigned int) cputime64_sub(cur_wall_time,
-                               j_dbs_info->prev_cpu_wall);
+               wall_time = (unsigned int)
+                       (cur_wall_time - j_dbs_info->prev_cpu_wall);
                j_dbs_info->prev_cpu_wall = cur_wall_time;
 
-               idle_time = (unsigned int) cputime64_sub(cur_idle_time,
-                               j_dbs_info->prev_cpu_idle);
+               idle_time = (unsigned int)
+                       (cur_idle_time - j_dbs_info->prev_cpu_idle);
                j_dbs_info->prev_cpu_idle = cur_idle_time;
 
-               iowait_time = (unsigned int) cputime64_sub(cur_iowait_time,
-                               j_dbs_info->prev_cpu_iowait);
+               iowait_time = (unsigned int)
+                       (cur_iowait_time - j_dbs_info->prev_cpu_iowait);
                j_dbs_info->prev_cpu_iowait = cur_iowait_time;
 
                if (dbs_tuners_ins.ignore_nice) {
index c5072a91e8489de1464ea2e1594c7faf7432f597..2a508edd768b9c24c67c45fd8fb86caeae3354c3 100644 (file)
@@ -61,9 +61,8 @@ static int cpufreq_stats_update(unsigned int cpu)
        spin_lock(&cpufreq_stats_lock);
        stat = per_cpu(cpufreq_stats_table, cpu);
        if (stat->time_in_state)
-               stat->time_in_state[stat->last_index] =
-                       cputime64_add(stat->time_in_state[stat->last_index],
-                                     cputime_sub(cur_time, stat->last_time));
+               stat->time_in_state[stat->last_index] +=
+                       cur_time - stat->last_time;
        stat->last_time = cur_time;
        spin_unlock(&cpufreq_stats_lock);
        return 0;
index 7a48b1eb423334780116f1cb6e16edcec8f19a49..5253d23361d91a4e93eefb00021bcd3ae4f9915f 100644 (file)
@@ -59,7 +59,7 @@ static ssize_t jz4740_hwmon_read_adcin(struct device *dev,
 {
        struct jz4740_hwmon *hwmon = dev_get_drvdata(dev);
        struct completion *completion = &hwmon->read_completion;
-       unsigned long t;
+       long t;
        unsigned long val;
        int ret;
 
@@ -203,7 +203,7 @@ static int __devexit jz4740_hwmon_remove(struct platform_device *pdev)
        return 0;
 }
 
-struct platform_driver jz4740_hwmon_driver = {
+static struct platform_driver jz4740_hwmon_driver = {
        .probe  = jz4740_hwmon_probe,
        .remove = __devexit_p(jz4740_hwmon_remove),
        .driver = {
index 66d7f1c7baa1e7d0fc742f384c694f805b95660f..6dc26b61219b423da10ec463df990532440c9ecc 100644 (file)
@@ -220,13 +220,11 @@ static void rackmeter_do_timer(struct work_struct *work)
        int i, offset, load, cumm, pause;
 
        cur_jiffies = jiffies64_to_cputime64(get_jiffies_64());
-       total_ticks = (unsigned int)cputime64_sub(cur_jiffies,
-                                                 rcpu->prev_wall);
+       total_ticks = (unsigned int) (cur_jiffies - rcpu->prev_wall);
        rcpu->prev_wall = cur_jiffies;
 
        total_idle_ticks = get_cpu_idle_time(cpu);
-       idle_ticks = (unsigned int) cputime64_sub(total_idle_ticks,
-                               rcpu->prev_idle);
+       idle_ticks = (unsigned int) (total_idle_ticks - rcpu->prev_idle);
        rcpu->prev_idle = total_idle_ticks;
 
        /* We do a very dumb calculation to update the LEDs for now,
index a1cb21f95302c497157dc97ed85f9acd68a3c10d..1e0e27cbe98786e82abf3d5d9e6ede36cc8f0eb2 100644 (file)
@@ -1606,6 +1606,14 @@ static const struct mmc_fixup blk_fixups[] =
                  MMC_QUIRK_BLK_NO_CMD23),
        MMC_FIXUP("MMC32G", 0x11, CID_OEMID_ANY, add_quirk_mmc,
                  MMC_QUIRK_BLK_NO_CMD23),
+
+       /*
+        * Some Micron MMC cards needs longer data read timeout than
+        * indicated in CSD.
+        */
+       MMC_FIXUP(CID_NAME_ANY, 0x13, 0x200, add_quirk_mmc,
+                 MMC_QUIRK_LONG_READ_TIME),
+
        END_FIXUP
 };
 
index 5278ffb20e74bbba9b5508e008171165ad296f16..950b97d7412a4f6f5f562401f132ed0d568a969a 100644 (file)
@@ -529,6 +529,18 @@ void mmc_set_data_timeout(struct mmc_data *data, const struct mmc_card *card)
                        data->timeout_clks = 0;
                }
        }
+
+       /*
+        * Some cards require longer data read timeout than indicated in CSD.
+        * Address this by setting the read timeout to a "reasonably high"
+        * value. For the cards tested, 300ms has proven enough. If necessary,
+        * this value can be increased if other problematic cards require this.
+        */
+       if (mmc_card_long_read_time(card) && data->flags & MMC_DATA_READ) {
+               data->timeout_ns = 300000000;
+               data->timeout_clks = 0;
+       }
+
        /*
         * Some cards need very high timeouts if driven in SPI mode.
         * The worst observed timeout was 900ms after writing a
@@ -1213,6 +1225,46 @@ void mmc_set_driver_type(struct mmc_host *host, unsigned int drv_type)
        mmc_host_clk_release(host);
 }
 
+static void mmc_poweroff_notify(struct mmc_host *host)
+{
+       struct mmc_card *card;
+       unsigned int timeout;
+       unsigned int notify_type = EXT_CSD_NO_POWER_NOTIFICATION;
+       int err = 0;
+
+       card = host->card;
+
+       /*
+        * Send power notify command only if card
+        * is mmc and notify state is powered ON
+        */
+       if (card && mmc_card_mmc(card) &&
+           (card->poweroff_notify_state == MMC_POWERED_ON)) {
+
+               if (host->power_notify_type == MMC_HOST_PW_NOTIFY_SHORT) {
+                       notify_type = EXT_CSD_POWER_OFF_SHORT;
+                       timeout = card->ext_csd.generic_cmd6_time;
+                       card->poweroff_notify_state = MMC_POWEROFF_SHORT;
+               } else {
+                       notify_type = EXT_CSD_POWER_OFF_LONG;
+                       timeout = card->ext_csd.power_off_longtime;
+                       card->poweroff_notify_state = MMC_POWEROFF_LONG;
+               }
+
+               err = mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
+                                EXT_CSD_POWER_OFF_NOTIFICATION,
+                                notify_type, timeout);
+
+               if (err && err != -EBADMSG)
+                       pr_err("Device failed to respond within %d poweroff "
+                              "time. Forcefully powering down the device\n",
+                              timeout);
+
+               /* Set the card state to no notification after the poweroff */
+               card->poweroff_notify_state = MMC_NO_POWER_NOTIFICATION;
+       }
+}
+
 /*
  * Apply power to the MMC stack.  This is a two-stage process.
  * First, we enable power to the card without the clock running.
@@ -1269,42 +1321,12 @@ static void mmc_power_up(struct mmc_host *host)
 
 void mmc_power_off(struct mmc_host *host)
 {
-       struct mmc_card *card;
-       unsigned int notify_type;
-       unsigned int timeout;
-       int err;
-
        mmc_host_clk_hold(host);
 
-       card = host->card;
        host->ios.clock = 0;
        host->ios.vdd = 0;
 
-       if (card && mmc_card_mmc(card) &&
-           (card->poweroff_notify_state == MMC_POWERED_ON)) {
-
-               if (host->power_notify_type == MMC_HOST_PW_NOTIFY_SHORT) {
-                       notify_type = EXT_CSD_POWER_OFF_SHORT;
-                       timeout = card->ext_csd.generic_cmd6_time;
-                       card->poweroff_notify_state = MMC_POWEROFF_SHORT;
-               } else {
-                       notify_type = EXT_CSD_POWER_OFF_LONG;
-                       timeout = card->ext_csd.power_off_longtime;
-                       card->poweroff_notify_state = MMC_POWEROFF_LONG;
-               }
-
-               err = mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
-                                EXT_CSD_POWER_OFF_NOTIFICATION,
-                                notify_type, timeout);
-
-               if (err && err != -EBADMSG)
-                       pr_err("Device failed to respond within %d poweroff "
-                              "time. Forcefully powering down the device\n",
-                              timeout);
-
-               /* Set the card state to no notification after the poweroff */
-               card->poweroff_notify_state = MMC_NO_POWER_NOTIFICATION;
-       }
+       mmc_poweroff_notify(host);
 
        /*
         * Reset ocr mask to be the highest possible voltage supported for
@@ -2196,7 +2218,7 @@ int mmc_card_sleep(struct mmc_host *host)
 
        mmc_bus_get(host);
 
-       if (host->bus_ops && !host->bus_dead && host->bus_ops->awake)
+       if (host->bus_ops && !host->bus_dead && host->bus_ops->sleep)
                err = host->bus_ops->sleep(host);
 
        mmc_bus_put(host);
@@ -2302,8 +2324,17 @@ int mmc_suspend_host(struct mmc_host *host)
                 * pre-claim the host.
                 */
                if (mmc_try_claim_host(host)) {
-                       if (host->bus_ops->suspend)
+                       if (host->bus_ops->suspend) {
+                               /*
+                                * For eMMC 4.5 device send notify command
+                                * before sleep, because in sleep state eMMC 4.5
+                                * devices respond to only RESET and AWAKE cmd
+                                */
+                               mmc_poweroff_notify(host);
                                err = host->bus_ops->suspend(host);
+                       }
+                       mmc_do_release_host(host);
+
                        if (err == -ENOSYS || !host->bus_ops->resume) {
                                /*
                                 * We simply "remove" the card in this case.
@@ -2318,7 +2349,6 @@ int mmc_suspend_host(struct mmc_host *host)
                                host->pm_flags = 0;
                                err = 0;
                        }
-                       mmc_do_release_host(host);
                } else {
                        err = -EBUSY;
                }
index dbf421a6279c702d91baeede5bb5f1e55dcb6fcd..d240427c12462dd545b233d6d6e75787e4b48abb 100644 (file)
@@ -876,17 +876,21 @@ static int mmc_init_card(struct mmc_host *host, u32 ocr,
         * set the notification byte in the ext_csd register of device
         */
        if ((host->caps2 & MMC_CAP2_POWEROFF_NOTIFY) &&
-           (card->poweroff_notify_state == MMC_NO_POWER_NOTIFICATION)) {
+           (card->ext_csd.rev >= 6)) {
                err = mmc_switch(card, EXT_CSD_CMD_SET_NORMAL,
                                 EXT_CSD_POWER_OFF_NOTIFICATION,
                                 EXT_CSD_POWER_ON,
                                 card->ext_csd.generic_cmd6_time);
                if (err && err != -EBADMSG)
                        goto free_card;
-       }
 
-       if (!err)
-               card->poweroff_notify_state = MMC_POWERED_ON;
+               /*
+                * The err can be -EBADMSG or 0,
+                * so check for success and update the flag
+                */
+               if (!err)
+                       card->poweroff_notify_state = MMC_POWERED_ON;
+       }
 
        /*
         * Activate high speed (if supported)
index 325ea61e12d37d4d47cb9f7ee82d795f11fc8ccf..8e0fbe99404778ee1cc4f512b767479d7dad9855 100644 (file)
@@ -732,6 +732,7 @@ static void mxcmci_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
                                "failed to config DMA channel. Falling back to PIO\n");
                        dma_release_channel(host->dma);
                        host->do_dma = 0;
+                       host->dma = NULL;
                }
        }
 
index 101cd31c82207982cf5986390a80f91c1053c535..d5fe43d53c51894ff9e861dbced02bfde10ef705 100644 (file)
@@ -1010,6 +1010,7 @@ static void omap_hsmmc_dma_cleanup(struct omap_hsmmc_host *host, int errno)
                        host->data->sg_len,
                        omap_hsmmc_get_dma_dir(host, host->data));
                omap_free_dma(dma_ch);
+               host->data->host_cookie = 0;
        }
        host->data = NULL;
 }
@@ -1575,8 +1576,10 @@ static void omap_hsmmc_post_req(struct mmc_host *mmc, struct mmc_request *mrq,
        struct mmc_data *data = mrq->data;
 
        if (host->use_dma) {
-               dma_unmap_sg(mmc_dev(host->mmc), data->sg, data->sg_len,
-                            omap_hsmmc_get_dma_dir(host, data));
+               if (data->host_cookie)
+                       dma_unmap_sg(mmc_dev(host->mmc), data->sg,
+                                    data->sg_len,
+                                    omap_hsmmc_get_dma_dir(host, data));
                data->host_cookie = 0;
        }
 }
index 4b920b7621cfe7cd214d7c3ed30d848de383e1ad..87b6f079b6e0e9bc367fd63ea363f47ca4c64eda 100644 (file)
@@ -15,6 +15,7 @@
 #include <linux/delay.h>
 #include <linux/device.h>
 #include <linux/mmc/host.h>
+#include <linux/module.h>
 #include <mach/cns3xxx.h>
 #include "sdhci-pltfm.h"
 
index 3d00e722efc9cec34e9cbdd1807863326f58e828..cb60c4197e0a2b805efd3fb6c9ebd95c5d7b4dc8 100644 (file)
@@ -644,8 +644,6 @@ static int sdhci_s3c_resume(struct platform_device *dev)
 static struct platform_driver sdhci_s3c_driver = {
        .probe          = sdhci_s3c_probe,
        .remove         = __devexit_p(sdhci_s3c_remove),
-       .suspend        = sdhci_s3c_suspend,
-       .resume         = sdhci_s3c_resume,
        .driver         = {
                .owner  = THIS_MODULE,
                .name   = "s3c-sdhci",
index 369366c8e205e4ed8aa27746e3a8940b89e54287..d5505f3fe2a170d04fc9011c5b5d2ff886a31df7 100644 (file)
@@ -908,7 +908,7 @@ static void sh_mmcif_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
                if (host->power) {
                        pm_runtime_put(&host->pd->dev);
                        host->power = false;
-                       if (p->down_pwr)
+                       if (p->down_pwr && ios->power_mode == MMC_POWER_OFF)
                                p->down_pwr(host->pd);
                }
                host->state = STATE_IDLE;
index d85a60cda16781ae25509a28705122597b87dd2a..4208b3958069119e68e8da8ad03e3eadf5862b8c 100644 (file)
@@ -798,7 +798,7 @@ static void tmio_mmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
                /* start bus clock */
                tmio_mmc_clk_start(host);
        } else if (ios->power_mode != MMC_POWER_UP) {
-               if (host->set_pwr)
+               if (host->set_pwr && ios->power_mode == MMC_POWER_OFF)
                        host->set_pwr(host->pdev, 0);
                if ((pdata->flags & TMIO_MMC_HAS_COLD_CD) &&
                    pdata->power) {
index 4144caf2f9d3a5ef95a7a9162b1de0cfa326cf4e..173b1d22e59b5a4bf8ed714f72b233cb70537468 100644 (file)
@@ -87,7 +87,7 @@ static int ceph_set_page_dirty(struct page *page)
        snapc = ceph_get_snap_context(ci->i_snap_realm->cached_context);
 
        /* dirty the head */
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (ci->i_head_snapc == NULL)
                ci->i_head_snapc = ceph_get_snap_context(snapc);
        ++ci->i_wrbuffer_ref_head;
@@ -100,7 +100,7 @@ static int ceph_set_page_dirty(struct page *page)
             ci->i_wrbuffer_ref-1, ci->i_wrbuffer_ref_head-1,
             ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head,
             snapc, snapc->seq, snapc->num_snaps);
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        /* now adjust page */
        spin_lock_irq(&mapping->tree_lock);
@@ -391,7 +391,7 @@ static struct ceph_snap_context *get_oldest_context(struct inode *inode,
        struct ceph_snap_context *snapc = NULL;
        struct ceph_cap_snap *capsnap = NULL;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
                dout(" cap_snap %p snapc %p has %d dirty pages\n", capsnap,
                     capsnap->context, capsnap->dirty_pages);
@@ -407,7 +407,7 @@ static struct ceph_snap_context *get_oldest_context(struct inode *inode,
                dout(" head snapc %p has %d dirty pages\n",
                     snapc, ci->i_wrbuffer_ref_head);
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return snapc;
 }
 
index 0f327c6c967954ec329f25bb44f45a1e4219cd71..8b53193e4f7ca67e3011c8501c8f1fee62357f9d 100644 (file)
@@ -309,7 +309,7 @@ void ceph_reservation_status(struct ceph_fs_client *fsc,
 /*
  * Find ceph_cap for given mds, if any.
  *
- * Called with i_lock held.
+ * Called with i_ceph_lock held.
  */
 static struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci, int mds)
 {
@@ -332,9 +332,9 @@ struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, int mds)
 {
        struct ceph_cap *cap;
 
-       spin_lock(&ci->vfs_inode.i_lock);
+       spin_lock(&ci->i_ceph_lock);
        cap = __get_cap_for_mds(ci, mds);
-       spin_unlock(&ci->vfs_inode.i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return cap;
 }
 
@@ -361,15 +361,16 @@ static int __ceph_get_cap_mds(struct ceph_inode_info *ci)
 
 int ceph_get_cap_mds(struct inode *inode)
 {
+       struct ceph_inode_info *ci = ceph_inode(inode);
        int mds;
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        mds = __ceph_get_cap_mds(ceph_inode(inode));
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return mds;
 }
 
 /*
- * Called under i_lock.
+ * Called under i_ceph_lock.
  */
 static void __insert_cap_node(struct ceph_inode_info *ci,
                              struct ceph_cap *new)
@@ -415,7 +416,7 @@ static void __cap_set_timeouts(struct ceph_mds_client *mdsc,
  *
  * If I_FLUSH is set, leave the inode at the front of the list.
  *
- * Caller holds i_lock
+ * Caller holds i_ceph_lock
  *    -> we take mdsc->cap_delay_lock
  */
 static void __cap_delay_requeue(struct ceph_mds_client *mdsc,
@@ -457,7 +458,7 @@ static void __cap_delay_requeue_front(struct ceph_mds_client *mdsc,
 /*
  * Cancel delayed work on cap.
  *
- * Caller must hold i_lock.
+ * Caller must hold i_ceph_lock.
  */
 static void __cap_delay_cancel(struct ceph_mds_client *mdsc,
                               struct ceph_inode_info *ci)
@@ -532,14 +533,14 @@ int ceph_add_cap(struct inode *inode,
                wanted |= ceph_caps_for_mode(fmode);
 
 retry:
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        cap = __get_cap_for_mds(ci, mds);
        if (!cap) {
                if (new_cap) {
                        cap = new_cap;
                        new_cap = NULL;
                } else {
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
                        new_cap = get_cap(mdsc, caps_reservation);
                        if (new_cap == NULL)
                                return -ENOMEM;
@@ -625,7 +626,7 @@ retry:
 
        if (fmode >= 0)
                __ceph_get_fmode(ci, fmode);
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        wake_up_all(&ci->i_cap_wq);
        return 0;
 }
@@ -792,7 +793,7 @@ int ceph_caps_revoking(struct ceph_inode_info *ci, int mask)
        struct rb_node *p;
        int ret = 0;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
                cap = rb_entry(p, struct ceph_cap, ci_node);
                if (__cap_is_valid(cap) &&
@@ -801,7 +802,7 @@ int ceph_caps_revoking(struct ceph_inode_info *ci, int mask)
                        break;
                }
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        dout("ceph_caps_revoking %p %s = %d\n", inode,
             ceph_cap_string(mask), ret);
        return ret;
@@ -855,7 +856,7 @@ int __ceph_caps_mds_wanted(struct ceph_inode_info *ci)
 }
 
 /*
- * called under i_lock
+ * called under i_ceph_lock
  */
 static int __ceph_is_any_caps(struct ceph_inode_info *ci)
 {
@@ -865,7 +866,7 @@ static int __ceph_is_any_caps(struct ceph_inode_info *ci)
 /*
  * Remove a cap.  Take steps to deal with a racing iterate_session_caps.
  *
- * caller should hold i_lock.
+ * caller should hold i_ceph_lock.
  * caller will not hold session s_mutex if called from destroy_inode.
  */
 void __ceph_remove_cap(struct ceph_cap *cap)
@@ -1028,7 +1029,7 @@ static void __queue_cap_release(struct ceph_mds_session *session,
 
 /*
  * Queue cap releases when an inode is dropped from our cache.  Since
- * inode is about to be destroyed, there is no need for i_lock.
+ * inode is about to be destroyed, there is no need for i_ceph_lock.
  */
 void ceph_queue_caps_release(struct inode *inode)
 {
@@ -1049,7 +1050,7 @@ void ceph_queue_caps_release(struct inode *inode)
 
 /*
  * Send a cap msg on the given inode.  Update our caps state, then
- * drop i_lock and send the message.
+ * drop i_ceph_lock and send the message.
  *
  * Make note of max_size reported/requested from mds, revoked caps
  * that have now been implemented.
@@ -1061,13 +1062,13 @@ void ceph_queue_caps_release(struct inode *inode)
  * Return non-zero if delayed release, or we experienced an error
  * such that the caller should requeue + retry later.
  *
- * called with i_lock, then drops it.
+ * called with i_ceph_lock, then drops it.
  * caller should hold snap_rwsem (read), s_mutex.
  */
 static int __send_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
                      int op, int used, int want, int retain, int flushing,
                      unsigned *pflush_tid)
-       __releases(cap->ci->vfs_inode->i_lock)
+       __releases(cap->ci->i_ceph_lock)
 {
        struct ceph_inode_info *ci = cap->ci;
        struct inode *inode = &ci->vfs_inode;
@@ -1170,7 +1171,7 @@ static int __send_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
                xattr_version = ci->i_xattrs.version;
        }
 
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        ret = send_cap_msg(session, ceph_vino(inode).ino, cap_id,
                op, keep, want, flushing, seq, flush_tid, issue_seq, mseq,
@@ -1198,13 +1199,13 @@ static int __send_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
  * Unless @again is true, skip cap_snaps that were already sent to
  * the MDS (i.e., during this session).
  *
- * Called under i_lock.  Takes s_mutex as needed.
+ * Called under i_ceph_lock.  Takes s_mutex as needed.
  */
 void __ceph_flush_snaps(struct ceph_inode_info *ci,
                        struct ceph_mds_session **psession,
                        int again)
-               __releases(ci->vfs_inode->i_lock)
-               __acquires(ci->vfs_inode->i_lock)
+               __releases(ci->i_ceph_lock)
+               __acquires(ci->i_ceph_lock)
 {
        struct inode *inode = &ci->vfs_inode;
        int mds;
@@ -1261,7 +1262,7 @@ retry:
                        session = NULL;
                }
                if (!session) {
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
                        mutex_lock(&mdsc->mutex);
                        session = __ceph_lookup_mds_session(mdsc, mds);
                        mutex_unlock(&mdsc->mutex);
@@ -1275,7 +1276,7 @@ retry:
                         * deletion or migration.  retry, and we'll
                         * get a better @mds value next time.
                         */
-                       spin_lock(&inode->i_lock);
+                       spin_lock(&ci->i_ceph_lock);
                        goto retry;
                }
 
@@ -1285,7 +1286,7 @@ retry:
                        list_del_init(&capsnap->flushing_item);
                list_add_tail(&capsnap->flushing_item,
                              &session->s_cap_snaps_flushing);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
 
                dout("flush_snaps %p cap_snap %p follows %lld tid %llu\n",
                     inode, capsnap, capsnap->follows, capsnap->flush_tid);
@@ -1302,7 +1303,7 @@ retry:
                next_follows = capsnap->follows + 1;
                ceph_put_cap_snap(capsnap);
 
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                goto retry;
        }
 
@@ -1322,11 +1323,9 @@ out:
 
 static void ceph_flush_snaps(struct ceph_inode_info *ci)
 {
-       struct inode *inode = &ci->vfs_inode;
-
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        __ceph_flush_snaps(ci, NULL, 0);
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 }
 
 /*
@@ -1373,7 +1372,7 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask)
  * Add dirty inode to the flushing list.  Assigned a seq number so we
  * can wait for caps to flush without starving.
  *
- * Called under i_lock.
+ * Called under i_ceph_lock.
  */
 static int __mark_caps_flushing(struct inode *inode,
                                 struct ceph_mds_session *session)
@@ -1421,9 +1420,9 @@ static int try_nonblocking_invalidate(struct inode *inode)
        struct ceph_inode_info *ci = ceph_inode(inode);
        u32 invalidating_gen = ci->i_rdcache_gen;
 
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        invalidate_mapping_pages(&inode->i_data, 0, -1);
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 
        if (inode->i_data.nrpages == 0 &&
            invalidating_gen == ci->i_rdcache_gen) {
@@ -1470,7 +1469,7 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags,
        if (mdsc->stopping)
                is_delayed = 1;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 
        if (ci->i_ceph_flags & CEPH_I_FLUSH)
                flags |= CHECK_CAPS_FLUSH;
@@ -1480,7 +1479,7 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags,
                __ceph_flush_snaps(ci, &session, 0);
        goto retry_locked;
 retry:
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 retry_locked:
        file_wanted = __ceph_caps_file_wanted(ci);
        used = __ceph_caps_used(ci);
@@ -1634,7 +1633,7 @@ ack:
                        if (mutex_trylock(&session->s_mutex) == 0) {
                                dout("inverting session/ino locks on %p\n",
                                     session);
-                               spin_unlock(&inode->i_lock);
+                               spin_unlock(&ci->i_ceph_lock);
                                if (took_snap_rwsem) {
                                        up_read(&mdsc->snap_rwsem);
                                        took_snap_rwsem = 0;
@@ -1648,7 +1647,7 @@ ack:
                        if (down_read_trylock(&mdsc->snap_rwsem) == 0) {
                                dout("inverting snap/in locks on %p\n",
                                     inode);
-                               spin_unlock(&inode->i_lock);
+                               spin_unlock(&ci->i_ceph_lock);
                                down_read(&mdsc->snap_rwsem);
                                took_snap_rwsem = 1;
                                goto retry;
@@ -1664,10 +1663,10 @@ ack:
                mds = cap->mds;  /* remember mds, so we don't repeat */
                sent++;
 
-               /* __send_cap drops i_lock */
+               /* __send_cap drops i_ceph_lock */
                delayed += __send_cap(mdsc, cap, CEPH_CAP_OP_UPDATE, used, want,
                                      retain, flushing, NULL);
-               goto retry; /* retake i_lock and restart our cap scan. */
+               goto retry; /* retake i_ceph_lock and restart our cap scan. */
        }
 
        /*
@@ -1681,7 +1680,7 @@ ack:
        else if (!is_delayed || force_requeue)
                __cap_delay_requeue(mdsc, ci);
 
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        if (queue_invalidate)
                ceph_queue_invalidate(inode);
@@ -1704,7 +1703,7 @@ static int try_flush_caps(struct inode *inode, struct ceph_mds_session *session,
        int flushing = 0;
 
 retry:
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (ci->i_ceph_flags & CEPH_I_NOFLUSH) {
                dout("try_flush_caps skipping %p I_NOFLUSH set\n", inode);
                goto out;
@@ -1716,7 +1715,7 @@ retry:
                int delayed;
 
                if (!session) {
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
                        session = cap->session;
                        mutex_lock(&session->s_mutex);
                        goto retry;
@@ -1727,18 +1726,18 @@ retry:
 
                flushing = __mark_caps_flushing(inode, session);
 
-               /* __send_cap drops i_lock */
+               /* __send_cap drops i_ceph_lock */
                delayed = __send_cap(mdsc, cap, CEPH_CAP_OP_FLUSH, used, want,
                                     cap->issued | cap->implemented, flushing,
                                     flush_tid);
                if (!delayed)
                        goto out_unlocked;
 
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                __cap_delay_requeue(mdsc, ci);
        }
 out:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 out_unlocked:
        if (session && unlock_session)
                mutex_unlock(&session->s_mutex);
@@ -1753,7 +1752,7 @@ static int caps_are_flushed(struct inode *inode, unsigned tid)
        struct ceph_inode_info *ci = ceph_inode(inode);
        int i, ret = 1;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        for (i = 0; i < CEPH_CAP_BITS; i++)
                if ((ci->i_flushing_caps & (1 << i)) &&
                    ci->i_cap_flush_tid[i] <= tid) {
@@ -1761,7 +1760,7 @@ static int caps_are_flushed(struct inode *inode, unsigned tid)
                        ret = 0;
                        break;
                }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return ret;
 }
 
@@ -1868,10 +1867,10 @@ int ceph_write_inode(struct inode *inode, struct writeback_control *wbc)
                struct ceph_mds_client *mdsc =
                        ceph_sb_to_client(inode->i_sb)->mdsc;
 
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                if (__ceph_caps_dirty(ci))
                        __cap_delay_requeue_front(mdsc, ci);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
        }
        return err;
 }
@@ -1894,7 +1893,7 @@ static void kick_flushing_capsnaps(struct ceph_mds_client *mdsc,
                struct inode *inode = &ci->vfs_inode;
                struct ceph_cap *cap;
 
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                cap = ci->i_auth_cap;
                if (cap && cap->session == session) {
                        dout("kick_flushing_caps %p cap %p capsnap %p\n", inode,
@@ -1904,7 +1903,7 @@ static void kick_flushing_capsnaps(struct ceph_mds_client *mdsc,
                        pr_err("%p auth cap %p not mds%d ???\n", inode,
                               cap, session->s_mds);
                }
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
        }
 }
 
@@ -1921,7 +1920,7 @@ void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
                struct ceph_cap *cap;
                int delayed = 0;
 
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                cap = ci->i_auth_cap;
                if (cap && cap->session == session) {
                        dout("kick_flushing_caps %p cap %p %s\n", inode,
@@ -1932,14 +1931,14 @@ void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
                                             cap->issued | cap->implemented,
                                             ci->i_flushing_caps, NULL);
                        if (delayed) {
-                               spin_lock(&inode->i_lock);
+                               spin_lock(&ci->i_ceph_lock);
                                __cap_delay_requeue(mdsc, ci);
-                               spin_unlock(&inode->i_lock);
+                               spin_unlock(&ci->i_ceph_lock);
                        }
                } else {
                        pr_err("%p auth cap %p not mds%d ???\n", inode,
                               cap, session->s_mds);
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
                }
        }
 }
@@ -1952,7 +1951,7 @@ static void kick_flushing_inode_caps(struct ceph_mds_client *mdsc,
        struct ceph_cap *cap;
        int delayed = 0;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        cap = ci->i_auth_cap;
        dout("kick_flushing_inode_caps %p flushing %s flush_seq %lld\n", inode,
             ceph_cap_string(ci->i_flushing_caps), ci->i_cap_flush_seq);
@@ -1964,12 +1963,12 @@ static void kick_flushing_inode_caps(struct ceph_mds_client *mdsc,
                                     cap->issued | cap->implemented,
                                     ci->i_flushing_caps, NULL);
                if (delayed) {
-                       spin_lock(&inode->i_lock);
+                       spin_lock(&ci->i_ceph_lock);
                        __cap_delay_requeue(mdsc, ci);
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
                }
        } else {
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
        }
 }
 
@@ -1978,7 +1977,7 @@ static void kick_flushing_inode_caps(struct ceph_mds_client *mdsc,
  * Take references to capabilities we hold, so that we don't release
  * them to the MDS prematurely.
  *
- * Protected by i_lock.
+ * Protected by i_ceph_lock.
  */
 static void __take_cap_refs(struct ceph_inode_info *ci, int got)
 {
@@ -2016,7 +2015,7 @@ static int try_get_cap_refs(struct ceph_inode_info *ci, int need, int want,
 
        dout("get_cap_refs %p need %s want %s\n", inode,
             ceph_cap_string(need), ceph_cap_string(want));
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 
        /* make sure file is actually open */
        file_wanted = __ceph_caps_file_wanted(ci);
@@ -2077,7 +2076,7 @@ static int try_get_cap_refs(struct ceph_inode_info *ci, int need, int want,
                     ceph_cap_string(have), ceph_cap_string(need));
        }
 out:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        dout("get_cap_refs %p ret %d got %s\n", inode,
             ret, ceph_cap_string(*got));
        return ret;
@@ -2094,7 +2093,7 @@ static void check_max_size(struct inode *inode, loff_t endoff)
        int check = 0;
 
        /* do we need to explicitly request a larger max_size? */
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if ((endoff >= ci->i_max_size ||
             endoff > (inode->i_size << 1)) &&
            endoff > ci->i_wanted_max_size) {
@@ -2103,7 +2102,7 @@ static void check_max_size(struct inode *inode, loff_t endoff)
                ci->i_wanted_max_size = endoff;
                check = 1;
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        if (check)
                ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL);
 }
@@ -2140,9 +2139,9 @@ retry:
  */
 void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps)
 {
-       spin_lock(&ci->vfs_inode.i_lock);
+       spin_lock(&ci->i_ceph_lock);
        __take_cap_refs(ci, caps);
-       spin_unlock(&ci->vfs_inode.i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 }
 
 /*
@@ -2160,7 +2159,7 @@ void ceph_put_cap_refs(struct ceph_inode_info *ci, int had)
        int last = 0, put = 0, flushsnaps = 0, wake = 0;
        struct ceph_cap_snap *capsnap;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (had & CEPH_CAP_PIN)
                --ci->i_pin_ref;
        if (had & CEPH_CAP_FILE_RD)
@@ -2193,7 +2192,7 @@ void ceph_put_cap_refs(struct ceph_inode_info *ci, int had)
                                }
                        }
                }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        dout("put_cap_refs %p had %s%s%s\n", inode, ceph_cap_string(had),
             last ? " last" : "", put ? " put" : "");
@@ -2225,7 +2224,7 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
        int found = 0;
        struct ceph_cap_snap *capsnap = NULL;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        ci->i_wrbuffer_ref -= nr;
        last = !ci->i_wrbuffer_ref;
 
@@ -2274,7 +2273,7 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
                }
        }
 
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        if (last) {
                ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL);
@@ -2291,7 +2290,7 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
  * Handle a cap GRANT message from the MDS.  (Note that a GRANT may
  * actually be a revocation if it specifies a smaller cap set.)
  *
- * caller holds s_mutex and i_lock, we drop both.
+ * caller holds s_mutex and i_ceph_lock, we drop both.
  *
  * return value:
  *  0 - ok
@@ -2302,7 +2301,7 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant,
                             struct ceph_mds_session *session,
                             struct ceph_cap *cap,
                             struct ceph_buffer *xattr_buf)
-               __releases(inode->i_lock)
+               __releases(ci->i_ceph_lock)
 {
        struct ceph_inode_info *ci = ceph_inode(inode);
        int mds = session->s_mds;
@@ -2453,7 +2452,7 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant,
        }
        BUG_ON(cap->issued & ~cap->implemented);
 
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        if (writeback)
                /*
                 * queue inode for writeback: we can't actually call
@@ -2483,7 +2482,7 @@ static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid,
                                 struct ceph_mds_caps *m,
                                 struct ceph_mds_session *session,
                                 struct ceph_cap *cap)
-       __releases(inode->i_lock)
+       __releases(ci->i_ceph_lock)
 {
        struct ceph_inode_info *ci = ceph_inode(inode);
        struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
@@ -2539,7 +2538,7 @@ static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid,
        wake_up_all(&ci->i_cap_wq);
 
 out:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        if (drop)
                iput(inode);
 }
@@ -2562,7 +2561,7 @@ static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
        dout("handle_cap_flushsnap_ack inode %p ci %p mds%d follows %lld\n",
             inode, ci, session->s_mds, follows);
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
                if (capsnap->follows == follows) {
                        if (capsnap->flush_tid != flush_tid) {
@@ -2585,7 +2584,7 @@ static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
                             capsnap, capsnap->follows);
                }
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        if (drop)
                iput(inode);
 }
@@ -2598,7 +2597,7 @@ static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
 static void handle_cap_trunc(struct inode *inode,
                             struct ceph_mds_caps *trunc,
                             struct ceph_mds_session *session)
-       __releases(inode->i_lock)
+       __releases(ci->i_ceph_lock)
 {
        struct ceph_inode_info *ci = ceph_inode(inode);
        int mds = session->s_mds;
@@ -2617,7 +2616,7 @@ static void handle_cap_trunc(struct inode *inode,
             inode, mds, seq, truncate_size, truncate_seq);
        queue_trunc = ceph_fill_file_size(inode, issued,
                                          truncate_seq, truncate_size, size);
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        if (queue_trunc)
                ceph_queue_vmtruncate(inode);
@@ -2646,7 +2645,7 @@ static void handle_cap_export(struct inode *inode, struct ceph_mds_caps *ex,
        dout("handle_cap_export inode %p ci %p mds%d mseq %d\n",
             inode, ci, mds, mseq);
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 
        /* make sure we haven't seen a higher mseq */
        for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
@@ -2690,7 +2689,7 @@ static void handle_cap_export(struct inode *inode, struct ceph_mds_caps *ex,
        }
        /* else, we already released it */
 
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 }
 
 /*
@@ -2745,9 +2744,9 @@ static void handle_cap_import(struct ceph_mds_client *mdsc,
        up_read(&mdsc->snap_rwsem);
 
        /* make sure we re-request max_size, if necessary */
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        ci->i_requested_max_size = 0;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 }
 
 /*
@@ -2762,6 +2761,7 @@ void ceph_handle_caps(struct ceph_mds_session *session,
        struct ceph_mds_client *mdsc = session->s_mdsc;
        struct super_block *sb = mdsc->fsc->sb;
        struct inode *inode;
+       struct ceph_inode_info *ci;
        struct ceph_cap *cap;
        struct ceph_mds_caps *h;
        int mds = session->s_mds;
@@ -2815,6 +2815,7 @@ void ceph_handle_caps(struct ceph_mds_session *session,
 
        /* lookup ino */
        inode = ceph_find_inode(sb, vino);
+       ci = ceph_inode(inode);
        dout(" op %s ino %llx.%llx inode %p\n", ceph_cap_op_name(op), vino.ino,
             vino.snap, inode);
        if (!inode) {
@@ -2844,16 +2845,16 @@ void ceph_handle_caps(struct ceph_mds_session *session,
        }
 
        /* the rest require a cap */
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        cap = __get_cap_for_mds(ceph_inode(inode), mds);
        if (!cap) {
                dout(" no cap on %p ino %llx.%llx from mds%d\n",
                     inode, ceph_ino(inode), ceph_snap(inode), mds);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                goto flush_cap_releases;
        }
 
-       /* note that each of these drops i_lock for us */
+       /* note that each of these drops i_ceph_lock for us */
        switch (op) {
        case CEPH_CAP_OP_REVOKE:
        case CEPH_CAP_OP_GRANT:
@@ -2869,7 +2870,7 @@ void ceph_handle_caps(struct ceph_mds_session *session,
                break;
 
        default:
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                pr_err("ceph_handle_caps: unknown cap op %d %s\n", op,
                       ceph_cap_op_name(op));
        }
@@ -2962,13 +2963,13 @@ void ceph_put_fmode(struct ceph_inode_info *ci, int fmode)
        struct inode *inode = &ci->vfs_inode;
        int last = 0;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        dout("put_fmode %p fmode %d %d -> %d\n", inode, fmode,
             ci->i_nr_by_mode[fmode], ci->i_nr_by_mode[fmode]-1);
        BUG_ON(ci->i_nr_by_mode[fmode] == 0);
        if (--ci->i_nr_by_mode[fmode] == 0)
                last++;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        if (last && ci->i_vino.snap == CEPH_NOSNAP)
                ceph_check_caps(ci, 0, NULL);
@@ -2991,7 +2992,7 @@ int ceph_encode_inode_release(void **p, struct inode *inode,
        int used, dirty;
        int ret = 0;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        used = __ceph_caps_used(ci);
        dirty = __ceph_caps_dirty(ci);
 
@@ -3046,7 +3047,7 @@ int ceph_encode_inode_release(void **p, struct inode *inode,
                             inode, cap, ceph_cap_string(cap->issued));
                }
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return ret;
 }
 
@@ -3061,7 +3062,7 @@ int ceph_encode_dentry_release(void **p, struct dentry *dentry,
 
        /*
         * force an record for the directory caps if we have a dentry lease.
-        * this is racy (can't take i_lock and d_lock together), but it
+        * this is racy (can't take i_ceph_lock and d_lock together), but it
         * doesn't have to be perfect; the mds will revoke anything we don't
         * release.
         */
index bca3948e9dbf6758746c896359471f4730407e24..3eeb976612625500b1a30b1fbc4896bdffa867c9 100644 (file)
@@ -281,18 +281,18 @@ static int ceph_readdir(struct file *filp, void *dirent, filldir_t filldir)
        }
 
        /* can we use the dcache? */
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if ((filp->f_pos == 2 || fi->dentry) &&
            !ceph_test_mount_opt(fsc, NOASYNCREADDIR) &&
            ceph_snap(inode) != CEPH_SNAPDIR &&
            ceph_dir_test_complete(inode) &&
            __ceph_caps_issued_mask(ci, CEPH_CAP_FILE_SHARED, 1)) {
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                err = __dcache_readdir(filp, dirent, filldir);
                if (err != -EAGAIN)
                        return err;
        } else {
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
        }
        if (fi->dentry) {
                err = note_last_dentry(fi, fi->dentry->d_name.name,
@@ -428,12 +428,12 @@ more:
         * were released during the whole readdir, and we should have
         * the complete dir contents in our cache.
         */
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (ci->i_release_count == fi->dir_release_count) {
                ceph_dir_set_complete(inode);
                ci->i_max_offset = filp->f_pos;
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        dout("readdir %p filp %p done.\n", inode, filp);
        return 0;
@@ -607,7 +607,7 @@ static struct dentry *ceph_lookup(struct inode *dir, struct dentry *dentry,
                struct ceph_inode_info *ci = ceph_inode(dir);
                struct ceph_dentry_info *di = ceph_dentry(dentry);
 
-               spin_lock(&dir->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                dout(" dir %p flags are %d\n", dir, ci->i_ceph_flags);
                if (strncmp(dentry->d_name.name,
                            fsc->mount_options->snapdir_name,
@@ -615,13 +615,13 @@ static struct dentry *ceph_lookup(struct inode *dir, struct dentry *dentry,
                    !is_root_ceph_dentry(dir, dentry) &&
                    ceph_dir_test_complete(dir) &&
                    (__ceph_caps_issued_mask(ci, CEPH_CAP_FILE_SHARED, 1))) {
-                       spin_unlock(&dir->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
                        dout(" dir %p complete, -ENOENT\n", dir);
                        d_add(dentry, NULL);
                        di->lease_shared_gen = ci->i_shared_gen;
                        return NULL;
                }
-               spin_unlock(&dir->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
        }
 
        op = ceph_snap(dir) == CEPH_SNAPDIR ?
@@ -841,12 +841,12 @@ static int drop_caps_for_unlink(struct inode *inode)
        struct ceph_inode_info *ci = ceph_inode(inode);
        int drop = CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (inode->i_nlink == 1) {
                drop |= ~(__ceph_caps_wanted(ci) | CEPH_CAP_PIN);
                ci->i_ceph_flags |= CEPH_I_NODELAY;
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return drop;
 }
 
@@ -1015,10 +1015,10 @@ static int dir_lease_is_valid(struct inode *dir, struct dentry *dentry)
        struct ceph_dentry_info *di = ceph_dentry(dentry);
        int valid = 0;
 
-       spin_lock(&dir->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (ci->i_shared_gen == di->lease_shared_gen)
                valid = __ceph_caps_issued_mask(ci, CEPH_CAP_FILE_SHARED, 1);
-       spin_unlock(&dir->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        dout("dir_lease_is_valid dir %p v%u dentry %p v%u = %d\n",
             dir, (unsigned)ci->i_shared_gen, dentry,
             (unsigned)di->lease_shared_gen, valid);
index ce549d31eeb7934634f4905e00861999cbe1b33f..ed72428d9c75c80a6744ccd6a996b83c1a20d333 100644 (file)
@@ -147,9 +147,9 @@ int ceph_open(struct inode *inode, struct file *file)
 
        /* trivially open snapdir */
        if (ceph_snap(inode) == CEPH_SNAPDIR) {
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                __ceph_get_fmode(ci, fmode);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                return ceph_init_file(inode, file, fmode);
        }
 
@@ -158,7 +158,7 @@ int ceph_open(struct inode *inode, struct file *file)
         * write) or any MDS (for read).  Update wanted set
         * asynchronously.
         */
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (__ceph_is_any_real_caps(ci) &&
            (((fmode & CEPH_FILE_MODE_WR) == 0) || ci->i_auth_cap)) {
                int mds_wanted = __ceph_caps_mds_wanted(ci);
@@ -168,7 +168,7 @@ int ceph_open(struct inode *inode, struct file *file)
                     inode, fmode, ceph_cap_string(wanted),
                     ceph_cap_string(issued));
                __ceph_get_fmode(ci, fmode);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
 
                /* adjust wanted? */
                if ((issued & wanted) != wanted &&
@@ -180,10 +180,10 @@ int ceph_open(struct inode *inode, struct file *file)
        } else if (ceph_snap(inode) != CEPH_NOSNAP &&
                   (ci->i_snap_caps & wanted) == wanted) {
                __ceph_get_fmode(ci, fmode);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                return ceph_init_file(inode, file, fmode);
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        dout("open fmode %d wants %s\n", fmode, ceph_cap_string(wanted));
        req = prepare_open_request(inode->i_sb, flags, 0);
@@ -743,9 +743,9 @@ retry_snap:
                 */
                int dirty;
 
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                ceph_put_cap_refs(ci, got);
 
                ret = generic_file_aio_write(iocb, iov, nr_segs, pos);
@@ -764,9 +764,9 @@ retry_snap:
 
        if (ret >= 0) {
                int dirty;
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                if (dirty)
                        __mark_inode_dirty(inode, dirty);
        }
@@ -797,7 +797,8 @@ static loff_t ceph_llseek(struct file *file, loff_t offset, int origin)
 
        mutex_lock(&inode->i_mutex);
        __ceph_do_pending_vmtruncate(inode);
-       if (origin != SEEK_CUR || origin != SEEK_SET) {
+
+       if (origin == SEEK_END || origin == SEEK_DATA || origin == SEEK_HOLE) {
                ret = ceph_do_getattr(inode, CEPH_STAT_CAP_SIZE);
                if (ret < 0) {
                        offset = ret;
index 116f36502f178f5ce0475bdfdd77073e9452e614..87fb132fb33012a9ca7839e9cca832eb472a1087 100644 (file)
@@ -297,6 +297,8 @@ struct inode *ceph_alloc_inode(struct super_block *sb)
 
        dout("alloc_inode %p\n", &ci->vfs_inode);
 
+       spin_lock_init(&ci->i_ceph_lock);
+
        ci->i_version = 0;
        ci->i_time_warp_seq = 0;
        ci->i_ceph_flags = 0;
@@ -583,7 +585,7 @@ static int fill_inode(struct inode *inode,
                               iinfo->xattr_len);
        }
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 
        /*
         * provided version will be odd if inode value is projected,
@@ -680,7 +682,7 @@ static int fill_inode(struct inode *inode,
                        char *sym;
 
                        BUG_ON(symlen != inode->i_size);
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
 
                        err = -ENOMEM;
                        sym = kmalloc(symlen+1, GFP_NOFS);
@@ -689,7 +691,7 @@ static int fill_inode(struct inode *inode,
                        memcpy(sym, iinfo->symlink, symlen);
                        sym[symlen] = 0;
 
-                       spin_lock(&inode->i_lock);
+                       spin_lock(&ci->i_ceph_lock);
                        if (!ci->i_symlink)
                                ci->i_symlink = sym;
                        else
@@ -715,7 +717,7 @@ static int fill_inode(struct inode *inode,
        }
 
 no_change:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        /* queue truncate if we saw i_size decrease */
        if (queue_trunc)
@@ -750,13 +752,13 @@ no_change:
                                     info->cap.flags,
                                     caps_reservation);
                } else {
-                       spin_lock(&inode->i_lock);
+                       spin_lock(&ci->i_ceph_lock);
                        dout(" %p got snap_caps %s\n", inode,
                             ceph_cap_string(le32_to_cpu(info->cap.caps)));
                        ci->i_snap_caps |= le32_to_cpu(info->cap.caps);
                        if (cap_fmode >= 0)
                                __ceph_get_fmode(ci, cap_fmode);
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
                }
        } else if (cap_fmode >= 0) {
                pr_warning("mds issued no caps on %llx.%llx\n",
@@ -849,19 +851,20 @@ static void ceph_set_dentry_offset(struct dentry *dn)
 {
        struct dentry *dir = dn->d_parent;
        struct inode *inode = dir->d_inode;
+       struct ceph_inode_info *ci = ceph_inode(inode);
        struct ceph_dentry_info *di;
 
        BUG_ON(!inode);
 
        di = ceph_dentry(dn);
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (!ceph_dir_test_complete(inode)) {
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                return;
        }
        di->offset = ceph_inode(inode)->i_max_offset++;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        spin_lock(&dir->d_lock);
        spin_lock_nested(&dn->d_lock, DENTRY_D_LOCK_NESTED);
@@ -1308,7 +1311,7 @@ int ceph_inode_set_size(struct inode *inode, loff_t size)
        struct ceph_inode_info *ci = ceph_inode(inode);
        int ret = 0;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        dout("set_size %p %llu -> %llu\n", inode, inode->i_size, size);
        inode->i_size = size;
        inode->i_blocks = (size + (1 << 9) - 1) >> 9;
@@ -1318,7 +1321,7 @@ int ceph_inode_set_size(struct inode *inode, loff_t size)
            (ci->i_reported_size << 1) < ci->i_max_size)
                ret = 1;
 
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return ret;
 }
 
@@ -1376,20 +1379,20 @@ static void ceph_invalidate_work(struct work_struct *work)
        u32 orig_gen;
        int check = 0;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        dout("invalidate_pages %p gen %d revoking %d\n", inode,
             ci->i_rdcache_gen, ci->i_rdcache_revoking);
        if (ci->i_rdcache_revoking != ci->i_rdcache_gen) {
                /* nevermind! */
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                goto out;
        }
        orig_gen = ci->i_rdcache_gen;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        truncate_inode_pages(&inode->i_data, 0);
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (orig_gen == ci->i_rdcache_gen &&
            orig_gen == ci->i_rdcache_revoking) {
                dout("invalidate_pages %p gen %d successful\n", inode,
@@ -1401,7 +1404,7 @@ static void ceph_invalidate_work(struct work_struct *work)
                     inode, orig_gen, ci->i_rdcache_gen,
                     ci->i_rdcache_revoking);
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        if (check)
                ceph_check_caps(ci, 0, NULL);
@@ -1460,10 +1463,10 @@ void __ceph_do_pending_vmtruncate(struct inode *inode)
        int wrbuffer_refs, wake = 0;
 
 retry:
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        if (ci->i_truncate_pending == 0) {
                dout("__do_pending_vmtruncate %p none pending\n", inode);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                return;
        }
 
@@ -1474,7 +1477,7 @@ retry:
        if (ci->i_wrbuffer_ref_head < ci->i_wrbuffer_ref) {
                dout("__do_pending_vmtruncate %p flushing snaps first\n",
                     inode);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                filemap_write_and_wait_range(&inode->i_data, 0,
                                             inode->i_sb->s_maxbytes);
                goto retry;
@@ -1484,15 +1487,15 @@ retry:
        wrbuffer_refs = ci->i_wrbuffer_ref;
        dout("__do_pending_vmtruncate %p (%d) to %lld\n", inode,
             ci->i_truncate_pending, to);
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        truncate_inode_pages(inode->i_mapping, to);
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        ci->i_truncate_pending--;
        if (ci->i_truncate_pending == 0)
                wake = 1;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        if (wrbuffer_refs == 0)
                ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL);
@@ -1547,7 +1550,7 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr)
        if (IS_ERR(req))
                return PTR_ERR(req);
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        issued = __ceph_caps_issued(ci, NULL);
        dout("setattr %p issued %s\n", inode, ceph_cap_string(issued));
 
@@ -1695,7 +1698,7 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr)
        }
 
        release &= issued;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        if (inode_dirty_flags)
                __mark_inode_dirty(inode, inode_dirty_flags);
@@ -1717,7 +1720,7 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr)
        __ceph_do_pending_vmtruncate(inode);
        return err;
 out:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        ceph_mdsc_put_request(req);
        return err;
 }
index 5a14c29cbba6f82b00ca42e8ac0327dbc12fd8bb..790914a598dd5d68b8f40b851c2faff2e790e4af 100644 (file)
@@ -241,11 +241,11 @@ static long ceph_ioctl_lazyio(struct file *file)
        struct ceph_inode_info *ci = ceph_inode(inode);
 
        if ((fi->fmode & CEPH_FILE_MODE_LAZY) == 0) {
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                ci->i_nr_by_mode[fi->fmode]--;
                fi->fmode |= CEPH_FILE_MODE_LAZY;
                ci->i_nr_by_mode[fi->fmode]++;
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                dout("ioctl_layzio: file %p marked lazy\n", file);
 
                ceph_check_caps(ci, 0, NULL);
index 264ab701154fead54aec35e0f45152e4709f201a..6203d805eb45061d20b5d8e08222f97aae6cd0a6 100644 (file)
@@ -732,21 +732,21 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
                }
        }
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        cap = NULL;
        if (mode == USE_AUTH_MDS)
                cap = ci->i_auth_cap;
        if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
                cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
        if (!cap) {
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                goto random;
        }
        mds = cap->session->s_mds;
        dout("choose_mds %p %llx.%llx mds%d (%scap %p)\n",
             inode, ceph_vinop(inode), mds,
             cap == ci->i_auth_cap ? "auth " : "", cap);
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return mds;
 
 random:
@@ -951,7 +951,7 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
 
        dout("removing cap %p, ci is %p, inode is %p\n",
             cap, ci, &ci->vfs_inode);
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        __ceph_remove_cap(cap);
        if (!__ceph_is_any_real_caps(ci)) {
                struct ceph_mds_client *mdsc =
@@ -984,7 +984,7 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
                }
                spin_unlock(&mdsc->cap_dirty_lock);
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        while (drop--)
                iput(inode);
        return 0;
@@ -1015,10 +1015,10 @@ static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
 
        wake_up_all(&ci->i_cap_wq);
        if (arg) {
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                ci->i_wanted_max_size = 0;
                ci->i_requested_max_size = 0;
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
        }
        return 0;
 }
@@ -1151,7 +1151,7 @@ static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
        if (session->s_trim_caps <= 0)
                return -1;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        mine = cap->issued | cap->implemented;
        used = __ceph_caps_used(ci);
        oissued = __ceph_caps_issued_other(ci, cap);
@@ -1170,7 +1170,7 @@ static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
                __ceph_remove_cap(cap);
        } else {
                /* try to drop referring dentries */
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                d_prune_aliases(inode);
                dout("trim_caps_cb %p cap %p  pruned, count now %d\n",
                     inode, cap, atomic_read(&inode->i_count));
@@ -1178,7 +1178,7 @@ static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
        }
 
 out:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return 0;
 }
 
@@ -1296,7 +1296,7 @@ static int check_cap_flush(struct ceph_mds_client *mdsc, u64 want_flush_seq)
                                           i_flushing_item);
                        struct inode *inode = &ci->vfs_inode;
 
-                       spin_lock(&inode->i_lock);
+                       spin_lock(&ci->i_ceph_lock);
                        if (ci->i_cap_flush_seq <= want_flush_seq) {
                                dout("check_cap_flush still flushing %p "
                                     "seq %lld <= %lld to mds%d\n", inode,
@@ -1304,7 +1304,7 @@ static int check_cap_flush(struct ceph_mds_client *mdsc, u64 want_flush_seq)
                                     session->s_mds);
                                ret = 0;
                        }
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
                }
                mutex_unlock(&session->s_mutex);
                ceph_put_mds_session(session);
@@ -1495,6 +1495,7 @@ retry:
                             pos, temp);
                } else if (stop_on_nosnap && inode &&
                           ceph_snap(inode) == CEPH_NOSNAP) {
+                       spin_unlock(&temp->d_lock);
                        break;
                } else {
                        pos -= temp->d_name.len;
@@ -2011,10 +2012,10 @@ void ceph_invalidate_dir_request(struct ceph_mds_request *req)
        struct ceph_inode_info *ci = ceph_inode(inode);
 
        dout("invalidate_dir_request %p (D_COMPLETE, lease(s))\n", inode);
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        ceph_dir_clear_complete(inode);
        ci->i_release_count++;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        if (req->r_dentry)
                ceph_invalidate_dentry_lease(req->r_dentry);
@@ -2422,7 +2423,7 @@ static int encode_caps_cb(struct inode *inode, struct ceph_cap *cap,
        if (err)
                goto out_free;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        cap->seq = 0;        /* reset cap seq */
        cap->issue_seq = 0;  /* and issue_seq */
 
@@ -2445,7 +2446,7 @@ static int encode_caps_cb(struct inode *inode, struct ceph_cap *cap,
                rec.v1.pathbase = cpu_to_le64(pathbase);
                reclen = sizeof(rec.v1);
        }
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 
        if (recon_state->flock) {
                int num_fcntl_locks, num_flock_locks;
index 4bb239921dbdf98a945b963d2d05ede9ff434736..a50ca0e39475794018c2350570547bff4f6a7df8 100644 (file)
@@ -20,7 +20,7 @@
  *
  *         mdsc->snap_rwsem
  *
- *         inode->i_lock
+ *         ci->i_ceph_lock
  *                 mdsc->snap_flush_lock
  *                 mdsc->cap_delay_lock
  *
index e2643719133323a07a8e69f0f631c306e80f7532..a559c80f127a04353a488181029744a009165f09 100644 (file)
@@ -446,7 +446,7 @@ void ceph_queue_cap_snap(struct ceph_inode_info *ci)
                return;
        }
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        used = __ceph_caps_used(ci);
        dirty = __ceph_caps_dirty(ci);
 
@@ -528,7 +528,7 @@ void ceph_queue_cap_snap(struct ceph_inode_info *ci)
                kfree(capsnap);
        }
 
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 }
 
 /*
@@ -537,7 +537,7 @@ void ceph_queue_cap_snap(struct ceph_inode_info *ci)
  *
  * If capsnap can now be flushed, add to snap_flush list, and return 1.
  *
- * Caller must hold i_lock.
+ * Caller must hold i_ceph_lock.
  */
 int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
                            struct ceph_cap_snap *capsnap)
@@ -739,9 +739,9 @@ static void flush_snaps(struct ceph_mds_client *mdsc)
                inode = &ci->vfs_inode;
                ihold(inode);
                spin_unlock(&mdsc->snap_flush_lock);
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                __ceph_flush_snaps(ci, &session, 0);
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                iput(inode);
                spin_lock(&mdsc->snap_flush_lock);
        }
@@ -847,7 +847,7 @@ void ceph_handle_snap(struct ceph_mds_client *mdsc,
                                continue;
                        ci = ceph_inode(inode);
 
-                       spin_lock(&inode->i_lock);
+                       spin_lock(&ci->i_ceph_lock);
                        if (!ci->i_snap_realm)
                                goto skip_inode;
                        /*
@@ -876,7 +876,7 @@ void ceph_handle_snap(struct ceph_mds_client *mdsc,
                        oldrealm = ci->i_snap_realm;
                        ci->i_snap_realm = realm;
                        spin_unlock(&realm->inodes_with_caps_lock);
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
 
                        ceph_get_snap_realm(mdsc, realm);
                        ceph_put_snap_realm(mdsc, oldrealm);
@@ -885,7 +885,7 @@ void ceph_handle_snap(struct ceph_mds_client *mdsc,
                        continue;
 
 skip_inode:
-                       spin_unlock(&inode->i_lock);
+                       spin_unlock(&ci->i_ceph_lock);
                        iput(inode);
                }
 
index 8dc73a594a90c11e31f20ef925104ea10e5b0b15..b48f15f101a0ed61127bc920b82afa3d61fcdc0b 100644 (file)
@@ -383,7 +383,7 @@ static int ceph_show_options(struct seq_file *m, struct vfsmount *mnt)
        if (fsopt->rsize != CEPH_RSIZE_DEFAULT)
                seq_printf(m, ",rsize=%d", fsopt->rsize);
        if (fsopt->rasize != CEPH_RASIZE_DEFAULT)
-               seq_printf(m, ",rasize=%d", fsopt->rsize);
+               seq_printf(m, ",rasize=%d", fsopt->rasize);
        if (fsopt->congestion_kb != default_congestion_kb())
                seq_printf(m, ",write_congestion_kb=%d", fsopt->congestion_kb);
        if (fsopt->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT)
index 01bf189e08a91387b6f4686b9999fe03ccf8aa32..edcbf3774a56460d377b31a9d51ea543a114a839 100644 (file)
@@ -220,7 +220,7 @@ struct ceph_dentry_info {
  * The locking for D_COMPLETE is a bit odd:
  *  - we can clear it at almost any time (see ceph_d_prune)
  *  - it is only meaningful if:
- *    - we hold dir inode i_lock
+ *    - we hold dir inode i_ceph_lock
  *    - we hold dir FILE_SHARED caps
  *    - the dentry D_COMPLETE is set
  */
@@ -250,6 +250,8 @@ struct ceph_inode_xattrs_info {
 struct ceph_inode_info {
        struct ceph_vino i_vino;   /* ceph ino + snap */
 
+       spinlock_t i_ceph_lock;
+
        u64 i_version;
        u32 i_time_warp_seq;
 
@@ -271,7 +273,7 @@ struct ceph_inode_info {
 
        struct ceph_inode_xattrs_info i_xattrs;
 
-       /* capabilities.  protected _both_ by i_lock and cap->session's
+       /* capabilities.  protected _both_ by i_ceph_lock and cap->session's
         * s_mutex. */
        struct rb_root i_caps;           /* cap list */
        struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
@@ -437,18 +439,18 @@ static inline void ceph_i_clear(struct inode *inode, unsigned mask)
 {
        struct ceph_inode_info *ci = ceph_inode(inode);
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        ci->i_ceph_flags &= ~mask;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 }
 
 static inline void ceph_i_set(struct inode *inode, unsigned mask)
 {
        struct ceph_inode_info *ci = ceph_inode(inode);
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        ci->i_ceph_flags |= mask;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
 }
 
 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
@@ -456,9 +458,9 @@ static inline bool ceph_i_test(struct inode *inode, unsigned mask)
        struct ceph_inode_info *ci = ceph_inode(inode);
        bool r;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        r = (ci->i_ceph_flags & mask) == mask;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return r;
 }
 
@@ -508,9 +510,9 @@ extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 {
        int issued;
-       spin_lock(&ci->vfs_inode.i_lock);
+       spin_lock(&ci->i_ceph_lock);
        issued = __ceph_caps_issued(ci, NULL);
-       spin_unlock(&ci->vfs_inode.i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return issued;
 }
 
@@ -518,9 +520,9 @@ static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
                                        int touch)
 {
        int r;
-       spin_lock(&ci->vfs_inode.i_lock);
+       spin_lock(&ci->i_ceph_lock);
        r = __ceph_caps_issued_mask(ci, mask, touch);
-       spin_unlock(&ci->vfs_inode.i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return r;
 }
 
@@ -743,10 +745,9 @@ extern int ceph_add_cap(struct inode *inode,
 extern void __ceph_remove_cap(struct ceph_cap *cap);
 static inline void ceph_remove_cap(struct ceph_cap *cap)
 {
-       struct inode *inode = &cap->ci->vfs_inode;
-       spin_lock(&inode->i_lock);
+       spin_lock(&cap->ci->i_ceph_lock);
        __ceph_remove_cap(cap);
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&cap->ci->i_ceph_lock);
 }
 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
                         struct ceph_cap *cap);
index 96c6739a02804f081adde4e2972523e1bf34911c..a5e36e4488a7d6e9f6dc0b9a5b9d440b4229e9c1 100644 (file)
@@ -343,8 +343,8 @@ void __ceph_destroy_xattrs(struct ceph_inode_info *ci)
 }
 
 static int __build_xattrs(struct inode *inode)
-       __releases(inode->i_lock)
-       __acquires(inode->i_lock)
+       __releases(ci->i_ceph_lock)
+       __acquires(ci->i_ceph_lock)
 {
        u32 namelen;
        u32 numattr = 0;
@@ -372,7 +372,7 @@ start:
                end = p + ci->i_xattrs.blob->vec.iov_len;
                ceph_decode_32_safe(&p, end, numattr, bad);
                xattr_version = ci->i_xattrs.version;
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
 
                xattrs = kcalloc(numattr, sizeof(struct ceph_xattr *),
                                 GFP_NOFS);
@@ -387,7 +387,7 @@ start:
                                goto bad_lock;
                }
 
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                if (ci->i_xattrs.version != xattr_version) {
                        /* lost a race, retry */
                        for (i = 0; i < numattr; i++)
@@ -418,7 +418,7 @@ start:
 
        return err;
 bad_lock:
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 bad:
        if (xattrs) {
                for (i = 0; i < numattr; i++)
@@ -512,7 +512,7 @@ ssize_t ceph_getxattr(struct dentry *dentry, const char *name, void *value,
        if (vxattrs)
                vxattr = ceph_match_vxattr(vxattrs, name);
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        dout("getxattr %p ver=%lld index_ver=%lld\n", inode,
             ci->i_xattrs.version, ci->i_xattrs.index_version);
 
@@ -520,14 +520,14 @@ ssize_t ceph_getxattr(struct dentry *dentry, const char *name, void *value,
            (ci->i_xattrs.index_version >= ci->i_xattrs.version)) {
                goto get_xattr;
        } else {
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                /* get xattrs from mds (if we don't already have them) */
                err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR);
                if (err)
                        return err;
        }
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 
        if (vxattr && vxattr->readonly) {
                err = vxattr->getxattr_cb(ci, value, size);
@@ -558,7 +558,7 @@ get_xattr:
        memcpy(value, xattr->val, xattr->val_len);
 
 out:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return err;
 }
 
@@ -573,7 +573,7 @@ ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
        u32 len;
        int i;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        dout("listxattr %p ver=%lld index_ver=%lld\n", inode,
             ci->i_xattrs.version, ci->i_xattrs.index_version);
 
@@ -581,13 +581,13 @@ ssize_t ceph_listxattr(struct dentry *dentry, char *names, size_t size)
            (ci->i_xattrs.index_version >= ci->i_xattrs.version)) {
                goto list_xattr;
        } else {
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                err = ceph_do_getattr(inode, CEPH_STAT_CAP_XATTR);
                if (err)
                        return err;
        }
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 
        err = __build_xattrs(inode);
        if (err < 0)
@@ -619,7 +619,7 @@ list_xattr:
                }
 
 out:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        return err;
 }
 
@@ -739,7 +739,7 @@ int ceph_setxattr(struct dentry *dentry, const char *name,
        if (!xattr)
                goto out;
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
 retry:
        issued = __ceph_caps_issued(ci, NULL);
        if (!(issued & CEPH_CAP_XATTR_EXCL))
@@ -752,12 +752,12 @@ retry:
            required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
                struct ceph_buffer *blob = NULL;
 
-               spin_unlock(&inode->i_lock);
+               spin_unlock(&ci->i_ceph_lock);
                dout(" preaallocating new blob size=%d\n", required_blob_size);
                blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
                if (!blob)
                        goto out;
-               spin_lock(&inode->i_lock);
+               spin_lock(&ci->i_ceph_lock);
                if (ci->i_xattrs.prealloc_blob)
                        ceph_buffer_put(ci->i_xattrs.prealloc_blob);
                ci->i_xattrs.prealloc_blob = blob;
@@ -770,13 +770,13 @@ retry:
        dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_XATTR_EXCL);
        ci->i_xattrs.dirty = true;
        inode->i_ctime = CURRENT_TIME;
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        if (dirty)
                __mark_inode_dirty(inode, dirty);
        return err;
 
 do_sync:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        err = ceph_sync_setxattr(dentry, name, value, size, flags);
 out:
        kfree(newname);
@@ -833,7 +833,7 @@ int ceph_removexattr(struct dentry *dentry, const char *name)
                        return -EOPNOTSUPP;
        }
 
-       spin_lock(&inode->i_lock);
+       spin_lock(&ci->i_ceph_lock);
        __build_xattrs(inode);
        issued = __ceph_caps_issued(ci, NULL);
        dout("removexattr %p issued %s\n", inode, ceph_cap_string(issued));
@@ -846,12 +846,12 @@ int ceph_removexattr(struct dentry *dentry, const char *name)
        ci->i_xattrs.dirty = true;
        inode->i_ctime = CURRENT_TIME;
 
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        if (dirty)
                __mark_inode_dirty(inode, dirty);
        return err;
 do_sync:
-       spin_unlock(&inode->i_lock);
+       spin_unlock(&ci->i_ceph_lock);
        err = ceph_send_removexattr(dentry, name);
        return err;
 }
index ca418aaf635254dd85722e791510f330c1a5917d..9d8715c45f2574ae473634feee7b0a07ca607355 100644 (file)
@@ -292,7 +292,7 @@ int __init configfs_inode_init(void)
        return bdi_init(&configfs_backing_dev_info);
 }
 
-void __exit configfs_inode_exit(void)
+void configfs_inode_exit(void)
 {
        bdi_destroy(&configfs_backing_dev_info);
 }
index ecc62178beda98d3975b2796714ab6d3d6b1968e..276e15cafd58e8b5171b8da6a8e503c99631cabd 100644 (file)
@@ -143,28 +143,26 @@ static int __init configfs_init(void)
                goto out;
 
        config_kobj = kobject_create_and_add("config", kernel_kobj);
-       if (!config_kobj) {
-               kmem_cache_destroy(configfs_dir_cachep);
-               configfs_dir_cachep = NULL;
-               goto out;
-       }
+       if (!config_kobj)
+               goto out2;
+
+       err = configfs_inode_init();
+       if (err)
+               goto out3;
 
        err = register_filesystem(&configfs_fs_type);
-       if (err) {
-               printk(KERN_ERR "configfs: Unable to register filesystem!\n");
-               kobject_put(config_kobj);
-               kmem_cache_destroy(configfs_dir_cachep);
-               configfs_dir_cachep = NULL;
-               goto out;
-       }
+       if (err)
+               goto out4;
 
-       err = configfs_inode_init();
-       if (err) {
-               unregister_filesystem(&configfs_fs_type);
-               kobject_put(config_kobj);
-               kmem_cache_destroy(configfs_dir_cachep);
-               configfs_dir_cachep = NULL;
-       }
+       return 0;
+out4:
+       printk(KERN_ERR "configfs: Unable to register filesystem!\n");
+       configfs_inode_exit();
+out3:
+       kobject_put(config_kobj);
+out2:
+       kmem_cache_destroy(configfs_dir_cachep);
+       configfs_dir_cachep = NULL;
 out:
        return err;
 }
index 61fa9e1614afd1922bae4cf5ce0d26dfdbab6b25..607b1557d292d1b24d5a605d28565de17f36bf13 100644 (file)
@@ -1095,7 +1095,7 @@ static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
                  le32_to_cpu(EXT_FIRST_INDEX(neh)->ei_block),
                  ext4_idx_pblock(EXT_FIRST_INDEX(neh)));
 
-       neh->eh_depth = cpu_to_le16(neh->eh_depth + 1);
+       neh->eh_depth = cpu_to_le16(le16_to_cpu(neh->eh_depth) + 1);
        ext4_mark_inode_dirty(handle, inode);
 out:
        brelse(bh);
@@ -2955,7 +2955,6 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
        /* Pre-conditions */
        BUG_ON(!ext4_ext_is_uninitialized(ex));
        BUG_ON(!in_range(map->m_lblk, ee_block, ee_len));
-       BUG_ON(map->m_lblk + map->m_len > ee_block + ee_len);
 
        /*
         * Attempt to transfer newly initialized blocks from the currently
index 848f436df29f6ffec6e8487549764d98c63d5b6b..92655fd8965737bac9aa002f173ed607ecdcfa8a 100644 (file)
@@ -1339,8 +1339,11 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd,
                                        clear_buffer_unwritten(bh);
                                }
 
-                               /* skip page if block allocation undone */
-                               if (buffer_delay(bh) || buffer_unwritten(bh))
+                               /*
+                                * skip page if block allocation undone and
+                                * block is dirty
+                                */
+                               if (ext4_bh_delay_or_unwritten(NULL, bh))
                                        skip_page = 1;
                                bh = bh->b_this_page;
                                block_start += bh->b_size;
@@ -2387,7 +2390,6 @@ static int ext4_da_write_begin(struct file *file, struct address_space *mapping,
        pgoff_t index;
        struct inode *inode = mapping->host;
        handle_t *handle;
-       loff_t page_len;
 
        index = pos >> PAGE_CACHE_SHIFT;
 
@@ -2434,13 +2436,6 @@ retry:
                 */
                if (pos + len > inode->i_size)
                        ext4_truncate_failed_write(inode);
-       } else {
-               page_len = pos & (PAGE_CACHE_SIZE - 1);
-               if (page_len > 0) {
-                       ret = ext4_discard_partial_page_buffers_no_lock(handle,
-                               inode, page, pos - page_len, page_len,
-                               EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED);
-               }
        }
 
        if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
@@ -2483,7 +2478,6 @@ static int ext4_da_write_end(struct file *file,
        loff_t new_i_size;
        unsigned long start, end;
        int write_mode = (int)(unsigned long)fsdata;
-       loff_t page_len;
 
        if (write_mode == FALL_BACK_TO_NONDELALLOC) {
                if (ext4_should_order_data(inode)) {
@@ -2508,7 +2502,7 @@ static int ext4_da_write_end(struct file *file,
         */
 
        new_i_size = pos + copied;
-       if (new_i_size > EXT4_I(inode)->i_disksize) {
+       if (copied && new_i_size > EXT4_I(inode)->i_disksize) {
                if (ext4_da_should_update_i_disksize(page, end)) {
                        down_write(&EXT4_I(inode)->i_data_sem);
                        if (new_i_size > EXT4_I(inode)->i_disksize) {
@@ -2532,16 +2526,6 @@ static int ext4_da_write_end(struct file *file,
        }
        ret2 = generic_write_end(file, mapping, pos, len, copied,
                                                        page, fsdata);
-
-       page_len = PAGE_CACHE_SIZE -
-                       ((pos + copied - 1) & (PAGE_CACHE_SIZE - 1));
-
-       if (page_len > 0) {
-               ret = ext4_discard_partial_page_buffers_no_lock(handle,
-                       inode, page, pos + copied - 1, page_len,
-                       EXT4_DISCARD_PARTIAL_PG_ZERO_UNMAPPED);
-       }
-
        copied = ret2;
        if (ret2 < 0)
                ret = ret2;
@@ -2781,10 +2765,11 @@ static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
                  iocb->private, io_end->inode->i_ino, iocb, offset,
                  size);
 
+       iocb->private = NULL;
+
        /* if not aio dio with unwritten extents, just free io and return */
        if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
                ext4_free_io_end(io_end);
-               iocb->private = NULL;
 out:
                if (is_async)
                        aio_complete(iocb, ret, 0);
@@ -2807,7 +2792,6 @@ out:
        spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
 
        /* queue the work to convert unwritten extents to written */
-       iocb->private = NULL;
        queue_work(wq, &io_end->work);
 
        /* XXX: probably should move into the real I/O completion handler */
@@ -3203,26 +3187,8 @@ int ext4_discard_partial_page_buffers_no_lock(handle_t *handle,
 
        iblock = index << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits);
 
-       if (!page_has_buffers(page)) {
-               /*
-                * If the range to be discarded covers a partial block
-                * we need to get the page buffers.  This is because
-                * partial blocks cannot be released and the page needs
-                * to be updated with the contents of the block before
-                * we write the zeros on top of it.
-                */
-               if ((from & (blocksize - 1)) ||
-                   ((from + length) & (blocksize - 1))) {
-                       create_empty_buffers(page, blocksize, 0);
-               } else {
-                       /*
-                        * If there are no partial blocks,
-                        * there is nothing to update,
-                        * so we can return now
-                        */
-                       return 0;
-               }
-       }
+       if (!page_has_buffers(page))
+               create_empty_buffers(page, blocksize, 0);
 
        /* Find the buffer that contains "offset" */
        bh = page_buffers(page);
index 7ce1d0b19c94576892d0d80f016f7db44110ccb7..7e106c810c62bb18435d9a0fcb6d5066a3813fad 100644 (file)
@@ -385,6 +385,18 @@ int ext4_bio_write_page(struct ext4_io_submit *io,
 
                block_end = block_start + blocksize;
                if (block_start >= len) {
+                       /*
+                        * Comments copied from block_write_full_page_endio:
+                        *
+                        * The page straddles i_size.  It must be zeroed out on
+                        * each and every writepage invocation because it may
+                        * be mmapped.  "A file is mapped in multiples of the
+                        * page size.  For a file that is not a multiple of
+                        * the  page size, the remaining memory is zeroed when
+                        * mapped, and writes to that region are not written
+                        * out to the file."
+                        */
+                       zero_user_segment(page, block_start, block_end);
                        clear_buffer_dirty(bh);
                        set_buffer_uptodate(bh);
                        continue;
index 3858767ec672ef214ad5f288f30932ee0b5d933a..3e1329e2f826132d8aec264116deb2c572be9ef7 100644 (file)
@@ -1155,9 +1155,9 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
                seq_puts(seq, ",block_validity");
 
        if (!test_opt(sb, INIT_INODE_TABLE))
-               seq_puts(seq, ",noinit_inode_table");
+               seq_puts(seq, ",noinit_itable");
        else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
-               seq_printf(seq, ",init_inode_table=%u",
+               seq_printf(seq, ",init_itable=%u",
                           (unsigned) sbi->s_li_wait_mult);
 
        ext4_show_quota_options(seq, sb);
@@ -1333,8 +1333,7 @@ enum {
        Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
        Opt_inode_readahead_blks, Opt_journal_ioprio,
        Opt_dioread_nolock, Opt_dioread_lock,
-       Opt_discard, Opt_nodiscard,
-       Opt_init_inode_table, Opt_noinit_inode_table,
+       Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
 };
 
 static const match_table_t tokens = {
@@ -1407,9 +1406,9 @@ static const match_table_t tokens = {
        {Opt_dioread_lock, "dioread_lock"},
        {Opt_discard, "discard"},
        {Opt_nodiscard, "nodiscard"},
-       {Opt_init_inode_table, "init_itable=%u"},
-       {Opt_init_inode_table, "init_itable"},
-       {Opt_noinit_inode_table, "noinit_itable"},
+       {Opt_init_itable, "init_itable=%u"},
+       {Opt_init_itable, "init_itable"},
+       {Opt_noinit_itable, "noinit_itable"},
        {Opt_err, NULL},
 };
 
@@ -1892,7 +1891,7 @@ set_qf_format:
                case Opt_dioread_lock:
                        clear_opt(sb, DIOREAD_NOLOCK);
                        break;
-               case Opt_init_inode_table:
+               case Opt_init_itable:
                        set_opt(sb, INIT_INODE_TABLE);
                        if (args[0].from) {
                                if (match_int(&args[0], &option))
@@ -1903,7 +1902,7 @@ set_qf_format:
                                return 0;
                        sbi->s_li_wait_mult = option;
                        break;
-               case Opt_noinit_inode_table:
+               case Opt_noinit_itable:
                        clear_opt(sb, INIT_INODE_TABLE);
                        break;
                default:
index 73c3992b2bb4aa765d3dc2418e907ca0a7ba78f2..ac86f8b3e3cb1cf050753148c3f3d065b18e6cf9 100644 (file)
@@ -156,6 +156,7 @@ __bdi_start_writeback(struct backing_dev_info *bdi, long nr_pages,
  * bdi_start_writeback - start writeback
  * @bdi: the backing device to write from
  * @nr_pages: the number of pages to write
+ * @reason: reason why some writeback work was initiated
  *
  * Description:
  *   This does WB_SYNC_NONE opportunistic writeback. The IO is only
@@ -1223,6 +1224,7 @@ static void wait_sb_inodes(struct super_block *sb)
  * writeback_inodes_sb_nr -    writeback dirty inodes from given super_block
  * @sb: the superblock
  * @nr: the number of pages to write
+ * @reason: reason why some writeback work initiated
  *
  * Start writeback on some inodes on this super_block. No guarantees are made
  * on how many (if any) will be written, and this function does not wait
@@ -1251,6 +1253,7 @@ EXPORT_SYMBOL(writeback_inodes_sb_nr);
 /**
  * writeback_inodes_sb -       writeback dirty inodes from given super_block
  * @sb: the superblock
+ * @reason: reason why some writeback work was initiated
  *
  * Start writeback on some inodes on this super_block. No guarantees are made
  * on how many (if any) will be written, and this function does not wait
@@ -1265,6 +1268,7 @@ EXPORT_SYMBOL(writeback_inodes_sb);
 /**
  * writeback_inodes_sb_if_idle -       start writeback if none underway
  * @sb: the superblock
+ * @reason: reason why some writeback work was initiated
  *
  * Invoke writeback_inodes_sb if no writeback is currently underway.
  * Returns 1 if writeback was started, 0 if not.
@@ -1285,6 +1289,7 @@ EXPORT_SYMBOL(writeback_inodes_sb_if_idle);
  * writeback_inodes_sb_if_idle -       start writeback if none underway
  * @sb: the superblock
  * @nr: the number of pages to write
+ * @reason: reason why some writeback work was initiated
  *
  * Invoke writeback_inodes_sb if no writeback is currently underway.
  * Returns 1 if writeback was started, 0 if not.
index 5cb8614508c339fb5e0c18031f3ebf249a4ec6c0..2aaf3eaaf13da03e328ac3523fe16ca09a60c690 100644 (file)
@@ -1512,7 +1512,7 @@ static int fuse_retrieve(struct fuse_conn *fc, struct inode *inode,
        else if (outarg->offset + num > file_size)
                num = file_size - outarg->offset;
 
-       while (num) {
+       while (num && req->num_pages < FUSE_MAX_PAGES_PER_REQ) {
                struct page *page;
                unsigned int this_num;
 
@@ -1526,6 +1526,7 @@ static int fuse_retrieve(struct fuse_conn *fc, struct inode *inode,
 
                num -= this_num;
                total_len += this_num;
+               index++;
        }
        req->misc.retrieve_in.offset = outarg->offset;
        req->misc.retrieve_in.size = total_len;
index 594f07a81c2899ba33a173be33cfc818afa0d39b..0c84100acd4492966e2fa48de6fbfacca0f10172 100644 (file)
@@ -1556,7 +1556,7 @@ static loff_t fuse_file_llseek(struct file *file, loff_t offset, int origin)
        struct inode *inode = file->f_path.dentry->d_inode;
 
        mutex_lock(&inode->i_mutex);
-       if (origin != SEEK_CUR || origin != SEEK_SET) {
+       if (origin != SEEK_CUR && origin != SEEK_SET) {
                retval = fuse_update_attributes(inode, NULL, file, NULL);
                if (retval)
                        goto exit;
@@ -1567,6 +1567,10 @@ static loff_t fuse_file_llseek(struct file *file, loff_t offset, int origin)
                offset += i_size_read(inode);
                break;
        case SEEK_CUR:
+               if (offset == 0) {
+                       retval = file->f_pos;
+                       goto exit;
+               }
                offset += file->f_pos;
                break;
        case SEEK_DATA:
index 3e6d727564792edd3b59dd6c509db3173e7a082f..aa83109b94316c9ed86bd0b4b5c30df5d9d9675f 100644 (file)
@@ -1138,28 +1138,28 @@ static int __init fuse_fs_init(void)
 {
        int err;
 
-       err = register_filesystem(&fuse_fs_type);
-       if (err)
-               goto out;
-
-       err = register_fuseblk();
-       if (err)
-               goto out_unreg;
-
        fuse_inode_cachep = kmem_cache_create("fuse_inode",
                                              sizeof(struct fuse_inode),
                                              0, SLAB_HWCACHE_ALIGN,
                                              fuse_inode_init_once);
        err = -ENOMEM;
        if (!fuse_inode_cachep)
-               goto out_unreg2;
+               goto out;
+
+       err = register_fuseblk();
+       if (err)
+               goto out2;
+
+       err = register_filesystem(&fuse_fs_type);
+       if (err)
+               goto out3;
 
        return 0;
 
- out_unreg2:
+ out3:
        unregister_fuseblk();
- out_unreg:
-       unregister_filesystem(&fuse_fs_type);
+ out2:
+       kmem_cache_destroy(fuse_inode_cachep);
  out:
        return err;
 }
index 5b5fa33b6b9dfd0384ca0cd7654363d8e2db560c..cbd1a61c110a38ca7a3e1d1e2de8b237e810b92e 100644 (file)
@@ -548,7 +548,7 @@ static int ncp_fill_super(struct super_block *sb, void *raw_data, int silent)
 
        error = bdi_setup_and_register(&server->bdi, "ncpfs", BDI_CAP_MAP_COPY);
        if (error)
-               goto out_bdi;
+               goto out_fput;
 
        server->ncp_filp = ncp_filp;
        server->ncp_sock = sock;
@@ -559,7 +559,7 @@ static int ncp_fill_super(struct super_block *sb, void *raw_data, int silent)
                error = -EBADF;
                server->info_filp = fget(data.info_fd);
                if (!server->info_filp)
-                       goto out_fput;
+                       goto out_bdi;
                error = -ENOTSOCK;
                sock_inode = server->info_filp->f_path.dentry->d_inode;
                if (!S_ISSOCK(sock_inode->i_mode))
@@ -746,9 +746,9 @@ out_nls:
 out_fput2:
        if (server->info_filp)
                fput(server->info_filp);
-out_fput:
-       bdi_destroy(&server->bdi);
 out_bdi:
+       bdi_destroy(&server->bdi);
+out_fput:
        /* 23/12/1998 Marcin Dalecki <dalecki@cs.net.pl>:
         * 
         * The previously used put_filp(ncp_filp); was bogus, since
index 3a1dafd228d14c5e11b142084305e8b60161f84c..8c344f037bd0195b8fad3a0630190cf32428883c 100644 (file)
@@ -394,8 +394,8 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 
        sigemptyset(&sigign);
        sigemptyset(&sigcatch);
-       cutime = cstime = utime = stime = cputime_zero;
-       cgtime = gtime = cputime_zero;
+       cutime = cstime = utime = stime = 0;
+       cgtime = gtime = 0;
 
        if (lock_task_sighand(task, &flags)) {
                struct signal_struct *sig = task->signal;
@@ -423,14 +423,14 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
                        do {
                                min_flt += t->min_flt;
                                maj_flt += t->maj_flt;
-                               gtime = cputime_add(gtime, t->gtime);
+                               gtime += t->gtime;
                                t = next_thread(t);
                        } while (t != task);
 
                        min_flt += sig->min_flt;
                        maj_flt += sig->maj_flt;
                        thread_group_times(task, &utime, &stime);
-                       gtime = cputime_add(gtime, sig->gtime);
+                       gtime += sig->gtime;
                }
 
                sid = task_session_nr_ns(task, ns);
index 9a8a2b77b87479621838ee00c928a43edbe938e9..03102d978180eba68469ef01d13ad25465f5b796 100644 (file)
@@ -91,20 +91,18 @@ static struct file_system_type proc_fs_type = {
 
 void __init proc_root_init(void)
 {
-       struct vfsmount *mnt;
        int err;
 
        proc_init_inodecache();
        err = register_filesystem(&proc_fs_type);
        if (err)
                return;
-       mnt = kern_mount_data(&proc_fs_type, &init_pid_ns);
-       if (IS_ERR(mnt)) {
+       err = pid_ns_prepare_proc(&init_pid_ns);
+       if (err) {
                unregister_filesystem(&proc_fs_type);
                return;
        }
 
-       init_pid_ns.proc_mnt = mnt;
        proc_symlink("mounts", NULL, "self/mounts");
 
        proc_net_init();
@@ -209,5 +207,5 @@ int pid_ns_prepare_proc(struct pid_namespace *ns)
 
 void pid_ns_release_proc(struct pid_namespace *ns)
 {
-       mntput(ns->proc_mnt);
+       kern_unmount(ns->proc_mnt);
 }
index 0fb22e464e725ed0fe154927c87ec67044c89adc..9610ac772d7e8e79a197725fbed6f77ef5b1dc96 100644 (file)
@@ -11,15 +11,20 @@ static int uptime_proc_show(struct seq_file *m, void *v)
 {
        struct timespec uptime;
        struct timespec idle;
+       u64 idletime;
+       u64 nsec;
+       u32 rem;
        int i;
-       u64 idletime = 0;
 
+       idletime = 0;
        for_each_possible_cpu(i)
-               idletime += kcpustat_cpu(i).cpustat[CPUTIME_IDLE];
+               idletime += (__force u64) kcpustat_cpu(i).cpustat[CPUTIME_IDLE];
 
        do_posix_clock_monotonic_gettime(&uptime);
        monotonic_to_bootbased(&uptime);
-       cputime_to_timespec(idletime, &idle);
+       nsec = cputime64_to_jiffies64(idletime) * TICK_NSEC;
+       idle.tv_sec = div_u64_rem(nsec, NSEC_PER_SEC, &rem);
+       idle.tv_nsec = rem;
        seq_printf(m, "%lu.%02lu %lu.%02lu\n",
                        (unsigned long) uptime.tv_sec,
                        (uptime.tv_nsec / (NSEC_PER_SEC / 100)),
index 20403dc5d4378da7a6e962601a8e60740cb6d5db..ae0e76bb6ebf44d5219693e27c08f9e9c8715da2 100644 (file)
@@ -2264,19 +2264,12 @@ static int __init ubifs_init(void)
                return -EINVAL;
        }
 
-       err = register_filesystem(&ubifs_fs_type);
-       if (err) {
-               ubifs_err("cannot register file system, error %d", err);
-               return err;
-       }
-
-       err = -ENOMEM;
        ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
                                sizeof(struct ubifs_inode), 0,
                                SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
                                &inode_slab_ctor);
        if (!ubifs_inode_slab)
-               goto out_reg;
+               return -ENOMEM;
 
        register_shrinker(&ubifs_shrinker_info);
 
@@ -2288,15 +2281,20 @@ static int __init ubifs_init(void)
        if (err)
                goto out_compr;
 
+       err = register_filesystem(&ubifs_fs_type);
+       if (err) {
+               ubifs_err("cannot register file system, error %d", err);
+               goto out_dbg;
+       }
        return 0;
 
+out_dbg:
+       dbg_debugfs_exit();
 out_compr:
        ubifs_compressors_exit();
 out_shrinker:
        unregister_shrinker(&ubifs_shrinker_info);
        kmem_cache_destroy(ubifs_inode_slab);
-out_reg:
-       unregister_filesystem(&ubifs_fs_type);
        return err;
 }
 /* late_initcall to let compressors initialize first */
index 62ce6823c0f2ac82ccdf285cf99dc8d36a9bb80b..77202e2c9fc508289d26aa4afb5f67ab26a9631b 100644 (file)
@@ -4,70 +4,64 @@
 #include <linux/time.h>
 #include <linux/jiffies.h>
 
-typedef unsigned long cputime_t;
+typedef unsigned long __nocast cputime_t;
 
-#define cputime_zero                   (0UL)
 #define cputime_one_jiffy              jiffies_to_cputime(1)
-#define cputime_max                    ((~0UL >> 1) - 1)
-#define cputime_add(__a, __b)          ((__a) +  (__b))
-#define cputime_sub(__a, __b)          ((__a) -  (__b))
-#define cputime_div(__a, __n)          ((__a) /  (__n))
-#define cputime_halve(__a)             ((__a) >> 1)
-#define cputime_eq(__a, __b)           ((__a) == (__b))
-#define cputime_gt(__a, __b)           ((__a) >  (__b))
-#define cputime_ge(__a, __b)           ((__a) >= (__b))
-#define cputime_lt(__a, __b)           ((__a) <  (__b))
-#define cputime_le(__a, __b)           ((__a) <= (__b))
-#define cputime_to_jiffies(__ct)       (__ct)
+#define cputime_to_jiffies(__ct)       (__force unsigned long)(__ct)
 #define cputime_to_scaled(__ct)                (__ct)
-#define jiffies_to_cputime(__hz)       (__hz)
+#define jiffies_to_cputime(__hz)       (__force cputime_t)(__hz)
 
-typedef u64 cputime64_t;
+typedef u64 __nocast cputime64_t;
 
-#define cputime64_zero (0ULL)
-#define cputime64_add(__a, __b)                ((__a) + (__b))
-#define cputime64_sub(__a, __b)                ((__a) - (__b))
-#define cputime64_to_jiffies64(__ct)   (__ct)
-#define jiffies64_to_cputime64(__jif)  (__jif)
-#define cputime_to_cputime64(__ct)     ((u64) __ct)
-#define cputime64_gt(__a, __b)         ((__a) >  (__b))
+#define cputime64_to_jiffies64(__ct)   (__force u64)(__ct)
+#define jiffies64_to_cputime64(__jif)  (__force cputime64_t)(__jif)
 
-#define nsecs_to_cputime64(__ct)       nsecs_to_jiffies64(__ct)
+#define nsecs_to_cputime64(__ct)       \
+       jiffies64_to_cputime64(nsecs_to_jiffies64(__ct))
 
 
 /*
  * Convert cputime to microseconds and back.
  */
-#define cputime_to_usecs(__ct)         jiffies_to_usecs(__ct)
-#define usecs_to_cputime(__msecs)      usecs_to_jiffies(__msecs)
+#define cputime_to_usecs(__ct)         \
+       jiffies_to_usecs(cputime_to_jiffies(__ct));
+#define usecs_to_cputime(__msecs)      \
+       jiffies_to_cputime(usecs_to_jiffies(__msecs));
 
 /*
  * Convert cputime to seconds and back.
  */
-#define cputime_to_secs(jif)           ((jif) / HZ)
-#define secs_to_cputime(sec)           ((sec) * HZ)
+#define cputime_to_secs(jif)           (cputime_to_jiffies(jif) / HZ)
+#define secs_to_cputime(sec)           jiffies_to_cputime((sec) * HZ)
 
 /*
  * Convert cputime to timespec and back.
  */
-#define timespec_to_cputime(__val)     timespec_to_jiffies(__val)
-#define cputime_to_timespec(__ct,__val)        jiffies_to_timespec(__ct,__val)
+#define timespec_to_cputime(__val)     \
+       jiffies_to_cputime(timespec_to_jiffies(__val))
+#define cputime_to_timespec(__ct,__val)        \
+       jiffies_to_timespec(cputime_to_jiffies(__ct),__val)
 
 /*
  * Convert cputime to timeval and back.
  */
-#define timeval_to_cputime(__val)      timeval_to_jiffies(__val)
-#define cputime_to_timeval(__ct,__val) jiffies_to_timeval(__ct,__val)
+#define timeval_to_cputime(__val)      \
+       jiffies_to_cputime(timeval_to_jiffies(__val))
+#define cputime_to_timeval(__ct,__val) \
+       jiffies_to_timeval(cputime_to_jiffies(__ct),__val)
 
 /*
  * Convert cputime to clock and back.
  */
-#define cputime_to_clock_t(__ct)       jiffies_to_clock_t(__ct)
-#define clock_t_to_cputime(__x)                clock_t_to_jiffies(__x)
+#define cputime_to_clock_t(__ct)       \
+       jiffies_to_clock_t(cputime_to_jiffies(__ct))
+#define clock_t_to_cputime(__x)                \
+       jiffies_to_cputime(clock_t_to_jiffies(__x))
 
 /*
  * Convert cputime64 to clock.
  */
-#define cputime64_to_clock_t(__ct)     jiffies_64_to_clock_t(__ct)
+#define cputime64_to_clock_t(__ct)     \
+       jiffies_64_to_clock_t(cputime64_to_jiffies64(__ct))
 
 #endif
index 4e4fbb820e204156729017b1488b77fe679ee30a..14b6cd022284dfec490a2119b06de2ecb9386b76 100644 (file)
        {0x1002, 0x6748, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6749, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6750, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_NEW_MEMMAP}, \
+       {0x1002, 0x6751, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6758, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6759, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_NEW_MEMMAP}, \
+       {0x1002, 0x675B, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_NEW_MEMMAP}, \
+       {0x1002, 0x675D, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x675F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6760, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CAICOS|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6761, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CAICOS|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6767, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CAICOS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6768, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CAICOS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6770, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CAICOS|RADEON_NEW_MEMMAP}, \
+       {0x1002, 0x6772, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CAICOS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6778, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CAICOS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6779, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CAICOS|RADEON_NEW_MEMMAP}, \
+       {0x1002, 0x677B, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CAICOS|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6840, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6841, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x6842, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TURKS|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x68f2, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CEDAR|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x68f8, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CEDAR|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x68f9, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CEDAR|RADEON_NEW_MEMMAP}, \
+       {0x1002, 0x68fa, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CEDAR|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x68fe, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CEDAR|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x7100, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_R520|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x7101, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_R520|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP}, \
        {0x1002, 0x9647, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_SUMO|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP|RADEON_IS_IGP},\
        {0x1002, 0x9648, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_SUMO|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP|RADEON_IS_IGP},\
        {0x1002, 0x964a, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_SUMO|RADEON_NEW_MEMMAP|RADEON_IS_IGP}, \
+       {0x1002, 0x964b, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_SUMO|RADEON_NEW_MEMMAP|RADEON_IS_IGP}, \
+       {0x1002, 0x964c, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_SUMO|RADEON_NEW_MEMMAP|RADEON_IS_IGP}, \
        {0x1002, 0x964e, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_SUMO|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP|RADEON_IS_IGP},\
        {0x1002, 0x964f, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_SUMO|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP|RADEON_IS_IGP},\
        {0x1002, 0x9710, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS880|RADEON_NEW_MEMMAP|RADEON_IS_IGP}, \
        {0x1002, 0x9805, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_PALM|RADEON_NEW_MEMMAP|RADEON_IS_IGP}, \
        {0x1002, 0x9806, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_PALM|RADEON_NEW_MEMMAP|RADEON_IS_IGP}, \
        {0x1002, 0x9807, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_PALM|RADEON_NEW_MEMMAP|RADEON_IS_IGP}, \
+       {0x1002, 0x9808, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_PALM|RADEON_NEW_MEMMAP|RADEON_IS_IGP}, \
+       {0x1002, 0x9809, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_PALM|RADEON_NEW_MEMMAP|RADEON_IS_IGP}, \
        {0, 0, 0}
 
 #define r128_PCI_IDS \
index 25b808631cd92c50d10cf6a31b2d9b9942b62ac9..fd7ff3d91e6a920ff084beca09d10b5b9abba981 100644 (file)
@@ -185,7 +185,6 @@ unsigned long __rounddown_pow_of_two(unsigned long n)
 #define rounddown_pow_of_two(n)                        \
 (                                              \
        __builtin_constant_p(n) ? (             \
-               (n == 1) ? 0 :                  \
                (1UL << ilog2(n))) :            \
        __rounddown_pow_of_two(n)               \
  )
index 415f2db414e17cd7c885b84e88b7b806746d3fe2..c8ef9bc54d500d0df052d4bd433e4b24d1855bb1 100644 (file)
@@ -218,6 +218,7 @@ struct mmc_card {
 #define MMC_QUIRK_INAND_CMD38  (1<<6)          /* iNAND devices have broken CMD38 */
 #define MMC_QUIRK_BLK_NO_CMD23 (1<<7)          /* Avoid CMD23 for regular multiblock */
 #define MMC_QUIRK_BROKEN_BYTE_MODE_512 (1<<8)  /* Avoid sending 512 bytes in */
+#define MMC_QUIRK_LONG_READ_TIME (1<<9)                /* Data read time > CSD says */
                                                /* byte mode */
        unsigned int    poweroff_notify_state;  /* eMMC4.5 notify feature */
 #define MMC_NO_POWER_NOTIFICATION      0
@@ -433,6 +434,11 @@ static inline int mmc_card_broken_byte_mode_512(const struct mmc_card *c)
        return c->quirks & MMC_QUIRK_BROKEN_BYTE_MODE_512;
 }
 
+static inline int mmc_card_long_read_time(const struct mmc_card *c)
+{
+       return c->quirks & MMC_QUIRK_LONG_READ_TIME;
+}
+
 #define mmc_card_name(c)       ((c)->cid.prod_name)
 #define mmc_card_id(c)         (dev_name(&(c)->dev))
 
index 5ef09012a629976b6e2ab6d3c248ff0a05dcafdd..5a2ab3c2757da6ce2bfa2153d42cf1041f828321 100644 (file)
@@ -485,8 +485,8 @@ struct task_cputime {
 
 #define INIT_CPUTIME   \
        (struct task_cputime) {                                 \
-               .utime = cputime_zero,                          \
-               .stime = cputime_zero,                          \
+               .utime = 0,                                     \
+               .stime = 0,                                     \
                .sum_exec_runtime = 0,                          \
        }
 
index 2e0ecfcc881dd124e3b2a2d868971fd1e6dea6f8..5b4293d9819d87b636d09d63224df43535e44fc4 100644 (file)
@@ -1269,7 +1269,7 @@ void mq_clear_sbinfo(struct ipc_namespace *ns)
 
 void mq_put_mnt(struct ipc_namespace *ns)
 {
-       mntput(ns->mq_mnt);
+       kern_unmount(ns->mq_mnt);
 }
 
 static int __init init_mqueue_fs(void)
@@ -1291,11 +1291,9 @@ static int __init init_mqueue_fs(void)
 
        spin_lock_init(&mq_lock);
 
-       init_ipc_ns.mq_mnt = kern_mount_data(&mqueue_fs_type, &init_ipc_ns);
-       if (IS_ERR(init_ipc_ns.mq_mnt)) {
-               error = PTR_ERR(init_ipc_ns.mq_mnt);
+       error = mq_init_ns(&init_ipc_ns);
+       if (error)
                goto out_filesystem;
-       }
 
        return 0;
 
index 8b5ce5d3f3ef3e4f468d5afc4175cd22518fc029..5652101cdac03d42ca9645017cc2a4f8df6537cb 100644 (file)
@@ -27,11 +27,6 @@ DEFINE_SPINLOCK(mq_lock);
  */
 struct ipc_namespace init_ipc_ns = {
        .count          = ATOMIC_INIT(1),
-#ifdef CONFIG_POSIX_MQUEUE
-       .mq_queues_max   = DFLT_QUEUESMAX,
-       .mq_msg_max      = DFLT_MSGMAX,
-       .mq_msgsize_max  = DFLT_MSGSIZEMAX,
-#endif
        .user_ns = &init_user_ns,
 };
 
index fa7eb3de2ddcd11eb5b0c7852aa840e9f4c19cc1..203dfead2e06de8e472d1a14b579462fc175dea9 100644 (file)
@@ -613,8 +613,8 @@ void acct_collect(long exitcode, int group_dead)
                pacct->ac_flag |= ACORE;
        if (current->flags & PF_SIGNALED)
                pacct->ac_flag |= AXSIG;
-       pacct->ac_utime = cputime_add(pacct->ac_utime, current->utime);
-       pacct->ac_stime = cputime_add(pacct->ac_stime, current->stime);
+       pacct->ac_utime += current->utime;
+       pacct->ac_stime += current->stime;
        pacct->ac_minflt += current->min_flt;
        pacct->ac_majflt += current->maj_flt;
        spin_unlock_irq(&current->sighand->siglock);
index 563f136094704fb8e6211636d72c3cc050da5f96..3f8ee8a138c4f7c53d0b16da77c1e8e3dc87b4d2 100644 (file)
@@ -178,8 +178,7 @@ static inline void check_for_tasks(int cpu)
        write_lock_irq(&tasklist_lock);
        for_each_process(p) {
                if (task_cpu(p) == cpu && p->state == TASK_RUNNING &&
-                   (!cputime_eq(p->utime, cputime_zero) ||
-                    !cputime_eq(p->stime, cputime_zero)))
+                   (p->utime || p->stime))
                        printk(KERN_WARNING "Task %s (pid = %d) is on cpu %d "
                                "(state = %ld, flags = %x)\n",
                                p->comm, task_pid_nr(p), cpu,
index d0b7d988f8735beb6e1e7b5f7b42ecbd04b91bf5..5e0d1f4c696efb6515f279392d6c323fea0739c5 100644 (file)
@@ -121,9 +121,9 @@ static void __exit_signal(struct task_struct *tsk)
                 * We won't ever get here for the group leader, since it
                 * will have been the last reference on the signal_struct.
                 */
-               sig->utime = cputime_add(sig->utime, tsk->utime);
-               sig->stime = cputime_add(sig->stime, tsk->stime);
-               sig->gtime = cputime_add(sig->gtime, tsk->gtime);
+               sig->utime += tsk->utime;
+               sig->stime += tsk->stime;
+               sig->gtime += tsk->gtime;
                sig->min_flt += tsk->min_flt;
                sig->maj_flt += tsk->maj_flt;
                sig->nvcsw += tsk->nvcsw;
@@ -1255,19 +1255,9 @@ static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
                spin_lock_irq(&p->real_parent->sighand->siglock);
                psig = p->real_parent->signal;
                sig = p->signal;
-               psig->cutime =
-                       cputime_add(psig->cutime,
-                       cputime_add(tgutime,
-                                   sig->cutime));
-               psig->cstime =
-                       cputime_add(psig->cstime,
-                       cputime_add(tgstime,
-                                   sig->cstime));
-               psig->cgtime =
-                       cputime_add(psig->cgtime,
-                       cputime_add(p->gtime,
-                       cputime_add(sig->gtime,
-                                   sig->cgtime)));
+               psig->cutime += tgutime + sig->cutime;
+               psig->cstime += tgstime + sig->cstime;
+               psig->cgtime += p->gtime + sig->gtime + sig->cgtime;
                psig->cmin_flt +=
                        p->min_flt + sig->min_flt + sig->cmin_flt;
                psig->cmaj_flt +=
index da4a6a10d088d2d575c4eeae9438f832efc6f41b..b058c5820ecd9f0dacaa66db83150e5f2510083f 100644 (file)
@@ -1023,8 +1023,8 @@ void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
  */
 static void posix_cpu_timers_init(struct task_struct *tsk)
 {
-       tsk->cputime_expires.prof_exp = cputime_zero;
-       tsk->cputime_expires.virt_exp = cputime_zero;
+       tsk->cputime_expires.prof_exp = 0;
+       tsk->cputime_expires.virt_exp = 0;
        tsk->cputime_expires.sched_exp = 0;
        INIT_LIST_HEAD(&tsk->cpu_timers[0]);
        INIT_LIST_HEAD(&tsk->cpu_timers[1]);
@@ -1132,14 +1132,10 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 
        init_sigpending(&p->pending);
 
-       p->utime = cputime_zero;
-       p->stime = cputime_zero;
-       p->gtime = cputime_zero;
-       p->utimescaled = cputime_zero;
-       p->stimescaled = cputime_zero;
+       p->utime = p->stime = p->gtime = 0;
+       p->utimescaled = p->stimescaled = 0;
 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
-       p->prev_utime = cputime_zero;
-       p->prev_stime = cputime_zero;
+       p->prev_utime = p->prev_stime = 0;
 #endif
 #if defined(SPLIT_RSS_COUNTING)
        memset(&p->rss_stat, 0, sizeof(p->rss_stat));
index d802883153da62210eea20e13f5f91170f67ebbf..22000c3db0dd53e28b01fc05a302316e02d1d221 100644 (file)
@@ -52,22 +52,22 @@ static void get_cpu_itimer(struct task_struct *tsk, unsigned int clock_id,
 
        cval = it->expires;
        cinterval = it->incr;
-       if (!cputime_eq(cval, cputime_zero)) {
+       if (cval) {
                struct task_cputime cputime;
                cputime_t t;
 
                thread_group_cputimer(tsk, &cputime);
                if (clock_id == CPUCLOCK_PROF)
-                       t = cputime_add(cputime.utime, cputime.stime);
+                       t = cputime.utime + cputime.stime;
                else
                        /* CPUCLOCK_VIRT */
                        t = cputime.utime;
 
-               if (cputime_le(cval, t))
+               if (cval < t)
                        /* about to fire */
                        cval = cputime_one_jiffy;
                else
-                       cval = cputime_sub(cval, t);
+                       cval = cval - t;
        }
 
        spin_unlock_irq(&tsk->sighand->siglock);
@@ -161,10 +161,9 @@ static void set_cpu_itimer(struct task_struct *tsk, unsigned int clock_id,
 
        cval = it->expires;
        cinterval = it->incr;
-       if (!cputime_eq(cval, cputime_zero) ||
-           !cputime_eq(nval, cputime_zero)) {
-               if (cputime_gt(nval, cputime_zero))
-                       nval = cputime_add(nval, cputime_one_jiffy);
+       if (cval || nval) {
+               if (nval > 0)
+                       nval += cputime_one_jiffy;
                set_process_cpu_timer(tsk, clock_id, &nval, &cval);
        }
        it->expires = nval;
index e7cb76dc18f51a9848c6300442f48653cdca28ca..125cb67daa214440868f1bcd332c35a4b4d18df9 100644 (file)
@@ -78,7 +78,7 @@ static inline int cpu_time_before(const clockid_t which_clock,
        if (CPUCLOCK_WHICH(which_clock) == CPUCLOCK_SCHED) {
                return now.sched < then.sched;
        }  else {
-               return cputime_lt(now.cpu, then.cpu);
+               return now.cpu < then.cpu;
        }
 }
 static inline void cpu_time_add(const clockid_t which_clock,
@@ -88,7 +88,7 @@ static inline void cpu_time_add(const clockid_t which_clock,
        if (CPUCLOCK_WHICH(which_clock) == CPUCLOCK_SCHED) {
                acc->sched += val.sched;
        }  else {
-               acc->cpu = cputime_add(acc->cpu, val.cpu);
+               acc->cpu += val.cpu;
        }
 }
 static inline union cpu_time_count cpu_time_sub(const clockid_t which_clock,
@@ -98,24 +98,11 @@ static inline union cpu_time_count cpu_time_sub(const clockid_t which_clock,
        if (CPUCLOCK_WHICH(which_clock) == CPUCLOCK_SCHED) {
                a.sched -= b.sched;
        }  else {
-               a.cpu = cputime_sub(a.cpu, b.cpu);
+               a.cpu -= b.cpu;
        }
        return a;
 }
 
-/*
- * Divide and limit the result to res >= 1
- *
- * This is necessary to prevent signal delivery starvation, when the result of
- * the division would be rounded down to 0.
- */
-static inline cputime_t cputime_div_non_zero(cputime_t time, unsigned long div)
-{
-       cputime_t res = cputime_div(time, div);
-
-       return max_t(cputime_t, res, 1);
-}
-
 /*
  * Update expiry time from increment, and increase overrun count,
  * given the current clock sample.
@@ -148,28 +135,26 @@ static void bump_cpu_timer(struct k_itimer *timer,
        } else {
                cputime_t delta, incr;
 
-               if (cputime_lt(now.cpu, timer->it.cpu.expires.cpu))
+               if (now.cpu < timer->it.cpu.expires.cpu)
                        return;
                incr = timer->it.cpu.incr.cpu;
-               delta = cputime_sub(cputime_add(now.cpu, incr),
-                                   timer->it.cpu.expires.cpu);
+               delta = now.cpu + incr - timer->it.cpu.expires.cpu;
                /* Don't use (incr*2 < delta), incr*2 might overflow. */
-               for (i = 0; cputime_lt(incr, cputime_sub(delta, incr)); i++)
-                            incr = cputime_add(incr, incr);
-               for (; i >= 0; incr = cputime_halve(incr), i--) {
-                       if (cputime_lt(delta, incr))
+               for (i = 0; incr < delta - incr; i++)
+                            incr += incr;
+               for (; i >= 0; incr = incr >> 1, i--) {
+                       if (delta < incr)
                                continue;
-                       timer->it.cpu.expires.cpu =
-                               cputime_add(timer->it.cpu.expires.cpu, incr);
+                       timer->it.cpu.expires.cpu += incr;
                        timer->it_overrun += 1 << i;
-                       delta = cputime_sub(delta, incr);
+                       delta -= incr;
                }
        }
 }
 
 static inline cputime_t prof_ticks(struct task_struct *p)
 {
-       return cputime_add(p->utime, p->stime);
+       return p->utime + p->stime;
 }
 static inline cputime_t virt_ticks(struct task_struct *p)
 {
@@ -248,8 +233,8 @@ void thread_group_cputime(struct task_struct *tsk, struct task_cputime *times)
 
        t = tsk;
        do {
-               times->utime = cputime_add(times->utime, t->utime);
-               times->stime = cputime_add(times->stime, t->stime);
+               times->utime += t->utime;
+               times->stime += t->stime;
                times->sum_exec_runtime += task_sched_runtime(t);
        } while_each_thread(tsk, t);
 out:
@@ -258,10 +243,10 @@ out:
 
 static void update_gt_cputime(struct task_cputime *a, struct task_cputime *b)
 {
-       if (cputime_gt(b->utime, a->utime))
+       if (b->utime > a->utime)
                a->utime = b->utime;
 
-       if (cputime_gt(b->stime, a->stime))
+       if (b->stime > a->stime)
                a->stime = b->stime;
 
        if (b->sum_exec_runtime > a->sum_exec_runtime)
@@ -306,7 +291,7 @@ static int cpu_clock_sample_group(const clockid_t which_clock,
                return -EINVAL;
        case CPUCLOCK_PROF:
                thread_group_cputime(p, &cputime);
-               cpu->cpu = cputime_add(cputime.utime, cputime.stime);
+               cpu->cpu = cputime.utime + cputime.stime;
                break;
        case CPUCLOCK_VIRT:
                thread_group_cputime(p, &cputime);
@@ -470,26 +455,24 @@ static void cleanup_timers(struct list_head *head,
                           unsigned long long sum_exec_runtime)
 {
        struct cpu_timer_list *timer, *next;
-       cputime_t ptime = cputime_add(utime, stime);
+       cputime_t ptime = utime + stime;
 
        list_for_each_entry_safe(timer, next, head, entry) {
                list_del_init(&timer->entry);
-               if (cputime_lt(timer->expires.cpu, ptime)) {
-                       timer->expires.cpu = cputime_zero;
+               if (timer->expires.cpu < ptime) {
+                       timer->expires.cpu = 0;
                } else {
-                       timer->expires.cpu = cputime_sub(timer->expires.cpu,
-                                                        ptime);
+                       timer->expires.cpu -= ptime;
                }
        }
 
        ++head;
        list_for_each_entry_safe(timer, next, head, entry) {
                list_del_init(&timer->entry);
-               if (cputime_lt(timer->expires.cpu, utime)) {
-                       timer->expires.cpu = cputime_zero;
+               if (timer->expires.cpu < utime) {
+                       timer->expires.cpu = 0;
                } else {
-                       timer->expires.cpu = cputime_sub(timer->expires.cpu,
-                                                        utime);
+                       timer->expires.cpu -= utime;
                }
        }
 
@@ -520,8 +503,7 @@ void posix_cpu_timers_exit_group(struct task_struct *tsk)
        struct signal_struct *const sig = tsk->signal;
 
        cleanup_timers(tsk->signal->cpu_timers,
-                      cputime_add(tsk->utime, sig->utime),
-                      cputime_add(tsk->stime, sig->stime),
+                      tsk->utime + sig->utime, tsk->stime + sig->stime,
                       tsk->se.sum_exec_runtime + sig->sum_sched_runtime);
 }
 
@@ -540,8 +522,7 @@ static void clear_dead_task(struct k_itimer *timer, union cpu_time_count now)
 
 static inline int expires_gt(cputime_t expires, cputime_t new_exp)
 {
-       return cputime_eq(expires, cputime_zero) ||
-              cputime_gt(expires, new_exp);
+       return expires == 0 || expires > new_exp;
 }
 
 /*
@@ -651,7 +632,7 @@ static int cpu_timer_sample_group(const clockid_t which_clock,
        default:
                return -EINVAL;
        case CPUCLOCK_PROF:
-               cpu->cpu = cputime_add(cputime.utime, cputime.stime);
+               cpu->cpu = cputime.utime + cputime.stime;
                break;
        case CPUCLOCK_VIRT:
                cpu->cpu = cputime.utime;
@@ -918,12 +899,12 @@ static void check_thread_timers(struct task_struct *tsk,
        unsigned long soft;
 
        maxfire = 20;
-       tsk->cputime_expires.prof_exp = cputime_zero;
+       tsk->cputime_expires.prof_exp = 0;
        while (!list_empty(timers)) {
                struct cpu_timer_list *t = list_first_entry(timers,
                                                      struct cpu_timer_list,
                                                      entry);
-               if (!--maxfire || cputime_lt(prof_ticks(tsk), t->expires.cpu)) {
+               if (!--maxfire || prof_ticks(tsk) < t->expires.cpu) {
                        tsk->cputime_expires.prof_exp = t->expires.cpu;
                        break;
                }
@@ -933,12 +914,12 @@ static void check_thread_timers(struct task_struct *tsk,
 
        ++timers;
        maxfire = 20;
-       tsk->cputime_expires.virt_exp = cputime_zero;
+       tsk->cputime_expires.virt_exp = 0;
        while (!list_empty(timers)) {
                struct cpu_timer_list *t = list_first_entry(timers,
                                                      struct cpu_timer_list,
                                                      entry);
-               if (!--maxfire || cputime_lt(virt_ticks(tsk), t->expires.cpu)) {
+               if (!--maxfire || virt_ticks(tsk) < t->expires.cpu) {
                        tsk->cputime_expires.virt_exp = t->expires.cpu;
                        break;
                }
@@ -1009,20 +990,19 @@ static u32 onecputick;
 static void check_cpu_itimer(struct task_struct *tsk, struct cpu_itimer *it,
                             cputime_t *expires, cputime_t cur_time, int signo)
 {
-       if (cputime_eq(it->expires, cputime_zero))
+       if (!it->expires)
                return;
 
-       if (cputime_ge(cur_time, it->expires)) {
-               if (!cputime_eq(it->incr, cputime_zero)) {
-                       it->expires = cputime_add(it->expires, it->incr);
+       if (cur_time >= it->expires) {
+               if (it->incr) {
+                       it->expires += it->incr;
                        it->error += it->incr_error;
                        if (it->error >= onecputick) {
-                               it->expires = cputime_sub(it->expires,
-                                                         cputime_one_jiffy);
+                               it->expires -= cputime_one_jiffy;
                                it->error -= onecputick;
                        }
                } else {
-                       it->expires = cputime_zero;
+                       it->expires = 0;
                }
 
                trace_itimer_expire(signo == SIGPROF ?
@@ -1031,9 +1011,7 @@ static void check_cpu_itimer(struct task_struct *tsk, struct cpu_itimer *it,
                __group_send_sig_info(signo, SEND_SIG_PRIV, tsk);
        }
 
-       if (!cputime_eq(it->expires, cputime_zero) &&
-           (cputime_eq(*expires, cputime_zero) ||
-            cputime_lt(it->expires, *expires))) {
+       if (it->expires && (!*expires || it->expires < *expires)) {
                *expires = it->expires;
        }
 }
@@ -1048,9 +1026,7 @@ static void check_cpu_itimer(struct task_struct *tsk, struct cpu_itimer *it,
  */
 static inline int task_cputime_zero(const struct task_cputime *cputime)
 {
-       if (cputime_eq(cputime->utime, cputime_zero) &&
-           cputime_eq(cputime->stime, cputime_zero) &&
-           cputime->sum_exec_runtime == 0)
+       if (!cputime->utime && !cputime->stime && !cputime->sum_exec_runtime)
                return 1;
        return 0;
 }
@@ -1076,15 +1052,15 @@ static void check_process_timers(struct task_struct *tsk,
         */
        thread_group_cputimer(tsk, &cputime);
        utime = cputime.utime;
-       ptime = cputime_add(utime, cputime.stime);
+       ptime = utime + cputime.stime;
        sum_sched_runtime = cputime.sum_exec_runtime;
        maxfire = 20;
-       prof_expires = cputime_zero;
+       prof_expires = 0;
        while (!list_empty(timers)) {
                struct cpu_timer_list *tl = list_first_entry(timers,
                                                      struct cpu_timer_list,
                                                      entry);
-               if (!--maxfire || cputime_lt(ptime, tl->expires.cpu)) {
+               if (!--maxfire || ptime < tl->expires.cpu) {
                        prof_expires = tl->expires.cpu;
                        break;
                }
@@ -1094,12 +1070,12 @@ static void check_process_timers(struct task_struct *tsk,
 
        ++timers;
        maxfire = 20;
-       virt_expires = cputime_zero;
+       virt_expires = 0;
        while (!list_empty(timers)) {
                struct cpu_timer_list *tl = list_first_entry(timers,
                                                      struct cpu_timer_list,
                                                      entry);
-               if (!--maxfire || cputime_lt(utime, tl->expires.cpu)) {
+               if (!--maxfire || utime < tl->expires.cpu) {
                        virt_expires = tl->expires.cpu;
                        break;
                }
@@ -1154,8 +1130,7 @@ static void check_process_timers(struct task_struct *tsk,
                        }
                }
                x = secs_to_cputime(soft);
-               if (cputime_eq(prof_expires, cputime_zero) ||
-                   cputime_lt(x, prof_expires)) {
+               if (!prof_expires || x < prof_expires) {
                        prof_expires = x;
                }
        }
@@ -1249,12 +1224,9 @@ out:
 static inline int task_cputime_expired(const struct task_cputime *sample,
                                        const struct task_cputime *expires)
 {
-       if (!cputime_eq(expires->utime, cputime_zero) &&
-           cputime_ge(sample->utime, expires->utime))
+       if (expires->utime && sample->utime >= expires->utime)
                return 1;
-       if (!cputime_eq(expires->stime, cputime_zero) &&
-           cputime_ge(cputime_add(sample->utime, sample->stime),
-                      expires->stime))
+       if (expires->stime && sample->utime + sample->stime >= expires->stime)
                return 1;
        if (expires->sum_exec_runtime != 0 &&
            sample->sum_exec_runtime >= expires->sum_exec_runtime)
@@ -1389,18 +1361,18 @@ void set_process_cpu_timer(struct task_struct *tsk, unsigned int clock_idx,
                 * it to be relative, *newval argument is relative and we update
                 * it to be absolute.
                 */
-               if (!cputime_eq(*oldval, cputime_zero)) {
-                       if (cputime_le(*oldval, now.cpu)) {
+               if (*oldval) {
+                       if (*oldval <= now.cpu) {
                                /* Just about to fire. */
                                *oldval = cputime_one_jiffy;
                        } else {
-                               *oldval = cputime_sub(*oldval, now.cpu);
+                               *oldval -= now.cpu;
                        }
                }
 
-               if (cputime_eq(*newval, cputime_zero))
+               if (!*newval)
                        return;
-               *newval = cputime_add(*newval, now.cpu);
+               *newval += now.cpu;
        }
 
        /*
index c7ea688faff4dd0000e0b9ad11e6c9110b1a98e6..cdf51a2adc2617449590c504745a5d4ebcd3de21 100644 (file)
@@ -935,7 +935,7 @@ static int irqtime_account_hi_update(void)
 
        local_irq_save(flags);
        latest_ns = this_cpu_read(cpu_hardirq_time);
-       if (cputime64_gt(nsecs_to_cputime64(latest_ns), cpustat[CPUTIME_IRQ]))
+       if (nsecs_to_cputime64(latest_ns) > cpustat[CPUTIME_IRQ])
                ret = 1;
        local_irq_restore(flags);
        return ret;
@@ -950,7 +950,7 @@ static int irqtime_account_si_update(void)
 
        local_irq_save(flags);
        latest_ns = this_cpu_read(cpu_softirq_time);
-       if (cputime64_gt(nsecs_to_cputime64(latest_ns), cpustat[CPUTIME_SOFTIRQ]))
+       if (nsecs_to_cputime64(latest_ns) > cpustat[CPUTIME_SOFTIRQ])
                ret = 1;
        local_irq_restore(flags);
        return ret;
@@ -2636,14 +2636,14 @@ void account_user_time(struct task_struct *p, cputime_t cputime,
        int index;
 
        /* Add user time to process. */
-       p->utime = cputime_add(p->utime, cputime);
-       p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
+       p->utime += cputime;
+       p->utimescaled += cputime_scaled;
        account_group_user_time(p, cputime);
 
        index = (TASK_NICE(p) > 0) ? CPUTIME_NICE : CPUTIME_USER;
 
        /* Add user time to cpustat. */
-       task_group_account_field(p, index, cputime);
+       task_group_account_field(p, index, (__force u64) cputime);
 
        /* Account for user time used */
        acct_update_integrals(p);
@@ -2658,24 +2658,21 @@ void account_user_time(struct task_struct *p, cputime_t cputime,
 static void account_guest_time(struct task_struct *p, cputime_t cputime,
                               cputime_t cputime_scaled)
 {
-       u64 tmp;
        u64 *cpustat = kcpustat_this_cpu->cpustat;
 
-       tmp = cputime_to_cputime64(cputime);
-
        /* Add guest time to process. */
-       p->utime = cputime_add(p->utime, cputime);
-       p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
+       p->utime += cputime;
+       p->utimescaled += cputime_scaled;
        account_group_user_time(p, cputime);
-       p->gtime = cputime_add(p->gtime, cputime);
+       p->gtime += cputime;
 
        /* Add guest time to cpustat. */
        if (TASK_NICE(p) > 0) {
-               cpustat[CPUTIME_NICE] += tmp;
-               cpustat[CPUTIME_GUEST_NICE] += tmp;
+               cpustat[CPUTIME_NICE] += (__force u64) cputime;
+               cpustat[CPUTIME_GUEST_NICE] += (__force u64) cputime;
        } else {
-               cpustat[CPUTIME_USER] += tmp;
-               cpustat[CPUTIME_GUEST] += tmp;
+               cpustat[CPUTIME_USER] += (__force u64) cputime;
+               cpustat[CPUTIME_GUEST] += (__force u64) cputime;
        }
 }
 
@@ -2691,12 +2688,12 @@ void __account_system_time(struct task_struct *p, cputime_t cputime,
                        cputime_t cputime_scaled, int index)
 {
        /* Add system time to process. */
-       p->stime = cputime_add(p->stime, cputime);
-       p->stimescaled = cputime_add(p->stimescaled, cputime_scaled);
+       p->stime += cputime;
+       p->stimescaled += cputime_scaled;
        account_group_system_time(p, cputime);
 
        /* Add system time to cpustat. */
-       task_group_account_field(p, index, cputime);
+       task_group_account_field(p, index, (__force u64) cputime);
 
        /* Account for system time used */
        acct_update_integrals(p);
@@ -2736,9 +2733,8 @@ void account_system_time(struct task_struct *p, int hardirq_offset,
 void account_steal_time(cputime_t cputime)
 {
        u64 *cpustat = kcpustat_this_cpu->cpustat;
-       u64 cputime64 = cputime_to_cputime64(cputime);
 
-       cpustat[CPUTIME_STEAL] += cputime64;
+       cpustat[CPUTIME_STEAL] += (__force u64) cputime;
 }
 
 /*
@@ -2748,13 +2744,12 @@ void account_steal_time(cputime_t cputime)
 void account_idle_time(cputime_t cputime)
 {
        u64 *cpustat = kcpustat_this_cpu->cpustat;
-       u64 cputime64 = cputime_to_cputime64(cputime);
        struct rq *rq = this_rq();
 
        if (atomic_read(&rq->nr_iowait) > 0)
-               cpustat[CPUTIME_IOWAIT] += cputime64;
+               cpustat[CPUTIME_IOWAIT] += (__force u64) cputime;
        else
-               cpustat[CPUTIME_IDLE] += cputime64;
+               cpustat[CPUTIME_IDLE] += (__force u64) cputime;
 }
 
 static __always_inline bool steal_account_process_tick(void)
@@ -2804,16 +2799,15 @@ static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
                                                struct rq *rq)
 {
        cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
-       u64 tmp = cputime_to_cputime64(cputime_one_jiffy);
        u64 *cpustat = kcpustat_this_cpu->cpustat;
 
        if (steal_account_process_tick())
                return;
 
        if (irqtime_account_hi_update()) {
-               cpustat[CPUTIME_IRQ] += tmp;
+               cpustat[CPUTIME_IRQ] += (__force u64) cputime_one_jiffy;
        } else if (irqtime_account_si_update()) {
-               cpustat[CPUTIME_SOFTIRQ] += tmp;
+               cpustat[CPUTIME_SOFTIRQ] += (__force u64) cputime_one_jiffy;
        } else if (this_cpu_ksoftirqd() == p) {
                /*
                 * ksoftirqd time do not get accounted in cpu_softirq_time.
@@ -2929,7 +2923,7 @@ void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
 
 void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
 {
-       cputime_t rtime, utime = p->utime, total = cputime_add(utime, p->stime);
+       cputime_t rtime, utime = p->utime, total = utime + p->stime;
 
        /*
         * Use CFS's precise accounting:
@@ -2937,11 +2931,11 @@ void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
        rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
 
        if (total) {
-               u64 temp = rtime;
+               u64 temp = (__force u64) rtime;
 
-               temp *= utime;
-               do_div(temp, total);
-               utime = (cputime_t)temp;
+               temp *= (__force u64) utime;
+               do_div(temp, (__force u32) total);
+               utime = (__force cputime_t) temp;
        } else
                utime = rtime;
 
@@ -2949,7 +2943,7 @@ void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
         * Compare with previous values, to keep monotonicity:
         */
        p->prev_utime = max(p->prev_utime, utime);
-       p->prev_stime = max(p->prev_stime, cputime_sub(rtime, p->prev_utime));
+       p->prev_stime = max(p->prev_stime, rtime - p->prev_utime);
 
        *ut = p->prev_utime;
        *st = p->prev_stime;
@@ -2966,21 +2960,20 @@ void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
 
        thread_group_cputime(p, &cputime);
 
-       total = cputime_add(cputime.utime, cputime.stime);
+       total = cputime.utime + cputime.stime;
        rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
 
        if (total) {
-               u64 temp = rtime;
+               u64 temp = (__force u64) rtime;
 
-               temp *= cputime.utime;
-               do_div(temp, total);
-               utime = (cputime_t)temp;
+               temp *= (__force u64) cputime.utime;
+               do_div(temp, (__force u32) total);
+               utime = (__force cputime_t) temp;
        } else
                utime = rtime;
 
        sig->prev_utime = max(sig->prev_utime, utime);
-       sig->prev_stime = max(sig->prev_stime,
-                             cputime_sub(rtime, sig->prev_utime));
+       sig->prev_stime = max(sig->prev_stime, rtime - sig->prev_utime);
 
        *ut = sig->prev_utime;
        *st = sig->prev_stime;
index ea2b6f0ec8688f3fe284a786ce548a2387fd168c..2ef90a51ec5e3ad6ccff9e8f44fde752d204dcda 100644 (file)
@@ -180,8 +180,7 @@ static inline void account_group_user_time(struct task_struct *tsk,
                return;
 
        raw_spin_lock(&cputimer->lock);
-       cputimer->cputime.utime =
-               cputime_add(cputimer->cputime.utime, cputime);
+       cputimer->cputime.utime += cputime;
        raw_spin_unlock(&cputimer->lock);
 }
 
@@ -204,8 +203,7 @@ static inline void account_group_system_time(struct task_struct *tsk,
                return;
 
        raw_spin_lock(&cputimer->lock);
-       cputimer->cputime.stime =
-               cputime_add(cputimer->cputime.stime, cputime);
+       cputimer->cputime.stime += cputime;
        raw_spin_unlock(&cputimer->lock);
 }
 
index b3f78d09a1053b67f62719a8879f6619ae12253b..739ef2bf105ccc03d1d5b115b92ca6dcef6ad18d 100644 (file)
@@ -1629,10 +1629,8 @@ bool do_notify_parent(struct task_struct *tsk, int sig)
        info.si_uid = __task_cred(tsk)->uid;
        rcu_read_unlock();
 
-       info.si_utime = cputime_to_clock_t(cputime_add(tsk->utime,
-                               tsk->signal->utime));
-       info.si_stime = cputime_to_clock_t(cputime_add(tsk->stime,
-                               tsk->signal->stime));
+       info.si_utime = cputime_to_clock_t(tsk->utime + tsk->signal->utime);
+       info.si_stime = cputime_to_clock_t(tsk->stime + tsk->signal->stime);
 
        info.si_status = tsk->exit_code & 0x7f;
        if (tsk->exit_code & 0x80)
index 481611fbd079a90094876bd2ef9743803fa45576..ddf8155bf3f8c09a3745a3cea82ec4e9aef49208 100644 (file)
@@ -1605,7 +1605,7 @@ static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
        unsigned long maxrss = 0;
 
        memset((char *) r, 0, sizeof *r);
-       utime = stime = cputime_zero;
+       utime = stime = 0;
 
        if (who == RUSAGE_THREAD) {
                task_times(current, &utime, &stime);
@@ -1635,8 +1635,8 @@ static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
 
                case RUSAGE_SELF:
                        thread_group_times(p, &tgutime, &tgstime);
-                       utime = cputime_add(utime, tgutime);
-                       stime = cputime_add(stime, tgstime);
+                       utime += tgutime;
+                       stime += tgstime;
                        r->ru_nvcsw += p->signal->nvcsw;
                        r->ru_nivcsw += p->signal->nivcsw;
                        r->ru_minflt += p->signal->min_flt;
index 5bbfac85866e7c333657da55e02637a135192ac8..23b4d784ebdd3e448ade6cb425e6ab3ba4e865fb 100644 (file)
@@ -127,7 +127,7 @@ void acct_update_integrals(struct task_struct *tsk)
 
                local_irq_save(flags);
                time = tsk->stime + tsk->utime;
-               dtime = cputime_sub(time, tsk->acct_timexpd);
+               dtime = time - tsk->acct_timexpd;
                jiffies_to_timeval(cputime_to_jiffies(dtime), &value);
                delta = value.tv_sec;
                delta = delta * USEC_PER_SEC + value.tv_usec;
index c0018f2d50e04e2ea03045989b742254be0a8489..c106d3b3cc640d56266c609f5c840e29c6fcae01 100644 (file)
@@ -2407,7 +2407,6 @@ static ssize_t generic_perform_write(struct file *file,
                                                iov_iter_count(i));
 
 again:
-
                /*
                 * Bring in the user page that we will copy from _first_.
                 * Otherwise there's a nasty deadlock on copying from the
@@ -2463,7 +2462,10 @@ again:
                written += copied;
 
                balance_dirty_pages_ratelimited(mapping);
-
+               if (fatal_signal_pending(current)) {
+                       status = -EINTR;
+                       break;
+               }
        } while (iov_iter_count(i));
 
        return written ? written : status;
index 71252486bc6f1f161b87592ded6e8b7e6c0dc8ee..50f08241f9815668d73b1adfcd5c9585f11bbc15 100644 (file)
@@ -411,8 +411,13 @@ void global_dirty_limits(unsigned long *pbackground, unsigned long *pdirty)
  *
  * Returns @bdi's dirty limit in pages. The term "dirty" in the context of
  * dirty balancing includes all PG_dirty, PG_writeback and NFS unstable pages.
- * And the "limit" in the name is not seriously taken as hard limit in
- * balance_dirty_pages().
+ *
+ * Note that balance_dirty_pages() will only seriously take it as a hard limit
+ * when sleeping max_pause per page is not enough to keep the dirty pages under
+ * control. For example, when the device is completely stalled due to some error
+ * conditions, or when there are 1000 dd tasks writing to a slow 10MB/s USB key.
+ * In the other normal situations, it acts more gently by throttling the tasks
+ * more (rather than completely block them) when the bdi dirty pages go high.
  *
  * It allocates high/low dirty limits to fast/slow devices, in order to prevent
  * - starving fast devices
@@ -594,6 +599,13 @@ static unsigned long bdi_position_ratio(struct backing_dev_info *bdi,
         */
        if (unlikely(bdi_thresh > thresh))
                bdi_thresh = thresh;
+       /*
+        * It's very possible that bdi_thresh is close to 0 not because the
+        * device is slow, but that it has remained inactive for long time.
+        * Honour such devices a reasonable good (hopefully IO efficient)
+        * threshold, so that the occasional writes won't be blocked and active
+        * writes can rampup the threshold quickly.
+        */
        bdi_thresh = max(bdi_thresh, (limit - dirty) / 8);
        /*
         * scale global setpoint to bdi's:
@@ -977,8 +989,7 @@ static unsigned long bdi_max_pause(struct backing_dev_info *bdi,
         *
         * 8 serves as the safety ratio.
         */
-       if (bdi_dirty)
-               t = min(t, bdi_dirty * HZ / (8 * bw + 1));
+       t = min(t, bdi_dirty * HZ / (8 * bw + 1));
 
        /*
         * The pause time will be settled within range (max_pause/4, max_pause).
@@ -1136,6 +1147,19 @@ pause:
                if (task_ratelimit)
                        break;
 
+               /*
+                * In the case of an unresponding NFS server and the NFS dirty
+                * pages exceeds dirty_thresh, give the other good bdi's a pipe
+                * to go through, so that tasks on them still remain responsive.
+                *
+                * In theory 1 page is enough to keep the comsumer-producer
+                * pipe going: the flusher cleans 1 page => the task dirties 1
+                * more page. However bdi_dirty has accounting errors.  So use
+                * the larger and more IO friendly bdi_stat_error.
+                */
+               if (bdi_dirty <= bdi_stat_error(bdi))
+                       break;
+
                if (fatal_signal_pending(current))
                        break;
        }
index 42599e31dcad8a6ceb1f3b29d9171f6a249e56a3..3a94eae7abe924d8d73d454b5a9330ab50bda968 100644 (file)
@@ -477,7 +477,6 @@ int crush_do_rule(struct crush_map *map,
        int i, j;
        int numrep;
        int firstn;
-       int rc = -1;
 
        BUG_ON(ruleno >= map->max_rules);
 
@@ -491,23 +490,18 @@ int crush_do_rule(struct crush_map *map,
         * that this may or may not correspond to the specific types
         * referenced by the crush rule.
         */
-       if (force >= 0) {
-               if (force >= map->max_devices ||
-                   map->device_parents[force] == 0) {
-                       /*dprintk("CRUSH: forcefed device dne\n");*/
-                       rc = -1;  /* force fed device dne */
-                       goto out;
-               }
-               if (!is_out(map, weight, force, x)) {
-                       while (1) {
-                               force_context[++force_pos] = force;
-                               if (force >= 0)
-                                       force = map->device_parents[force];
-                               else
-                                       force = map->bucket_parents[-1-force];
-                               if (force == 0)
-                                       break;
-                       }
+       if (force >= 0 &&
+           force < map->max_devices &&
+           map->device_parents[force] != 0 &&
+           !is_out(map, weight, force, x)) {
+               while (1) {
+                       force_context[++force_pos] = force;
+                       if (force >= 0)
+                               force = map->device_parents[force];
+                       else
+                               force = map->bucket_parents[-1-force];
+                       if (force == 0)
+                               break;
                }
        }
 
@@ -600,10 +594,7 @@ int crush_do_rule(struct crush_map *map,
                        BUG_ON(1);
                }
        }
-       rc = result_len;
-
-out:
-       return rc;
+       return result_len;
 }
 
 
This page took 0.120836 seconds and 5 git commands to generate.