deliverable/linux.git
16 years agoocfs2: Add xattr bucket iteration for large numbers of EAs
Tao Ma [Mon, 18 Aug 2008 09:38:50 +0000 (17:38 +0800)] 
ocfs2: Add xattr bucket iteration for large numbers of EAs

Ocfs2 breaks up xattr index tree leaves into 4k regions, called buckets.
Attributes are stored within a given bucket, depending on hash value.

After a discussion with Mark, we decided that the per-bucket index
(xe_entry[]) would only exist in the 1st block of a bucket. Likewise,
name/value pairs will not straddle more than one block. This allows the
majority of operations to work directly on the buffer heads in a leaf block.

This patch adds code to iterate the buckets in an EA. A new abstration of
ocfs2_xattr_bucket is added. It records the bhs in this bucket and
ocfs2_xattr_header. This keeps the code neat, improving readibility.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Add xattr index tree operations
Tao Ma [Mon, 18 Aug 2008 09:38:49 +0000 (17:38 +0800)] 
ocfs2: Add xattr index tree operations

When necessary, an ocfs2_xattr_block will embed an ocfs2_extent_list to
store large numbers of EAs. This patch adds a new type in
ocfs2_extent_tree_type and adds the implementation so that we can re-use the
b-tree code to handle the storage of many EAs.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Add extended attribute support
Tiger Yang [Mon, 18 Aug 2008 09:11:00 +0000 (17:11 +0800)] 
ocfs2: Add extended attribute support

This patch implements storing extended attributes both in inode or a single
external block. We only store EA's in-inode when blocksize > 512 or that
inode block has free space for it. When an EA's value is larger than 80
bytes, we will store the value via b-tree outside inode or block.

Signed-off-by: Tiger Yang <tiger.yang@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: reserve inline space for extended attribute
Tiger Yang [Mon, 18 Aug 2008 09:08:55 +0000 (17:08 +0800)] 
ocfs2: reserve inline space for extended attribute

Add the structures and helper functions we want for handling inline extended
attributes. We also update the inline-data handlers so that they properly
function in the event that we have both inline data and inline attributes
sharing an inode block.

Signed-off-by: Tiger Yang <tiger.yang@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Add extent tree operation for xattr value btrees
Tao Ma [Mon, 18 Aug 2008 09:38:48 +0000 (17:38 +0800)] 
ocfs2: Add extent tree operation for xattr value btrees

Add some thin wrappers around ocfs2_insert_extent() for each of the 3
different btree types, ocfs2_inode_insert_extent(),
ocfs2_xattr_value_insert_extent() and ocfs2_xattr_tree_insert_extent(). The
last is for the xattr index btree, which will be used in a followup patch.

All the old callers in file.c etc will call ocfs2_dinode_insert_extent(),
while the other two handle the xattr issue. And the init of extent tree are
handled by these functions.

When storing xattr value which is too large, we will allocate some clusters
for it and here ocfs2_extent_list and ocfs2_extent_rec will also be used. In
order to re-use the b-tree operation code, a new parameter named "private"
is added into ocfs2_extent_tree and it is used to indicate the root of
ocfs2_exent_list. The reason is that we can't deduce the root from the
buffer_head now. It may be in an inode, an ocfs2_xattr_block or even worse,
in any place in an ocfs2_xattr_bucket.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Add helper function in uptodate.c for removing xattr clusters
Tao Ma [Mon, 18 Aug 2008 09:38:47 +0000 (17:38 +0800)] 
ocfs2: Add helper function in uptodate.c for removing xattr clusters

The old uptodate only handles the issue of removing one buffer_head from
ocfs2 inode's buffer cache. With xattr clusters, we may need to remove
multiple buffer_head's at a time.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Add the basic xattr disk layout in ocfs2_fs.h
Tao Ma [Mon, 18 Aug 2008 09:38:46 +0000 (17:38 +0800)] 
ocfs2: Add the basic xattr disk layout in ocfs2_fs.h

Ocfs2 uses a very flexible structure for storing extended attributes on
disk. Small amount of attributes are stored directly in the inode block - up
to 256 bytes worth. If that fills up, attributes are also stored in an
external block, linked to from the inode block. That block can in turn
expand to a btree, capable of storing large numbers of attributes.

Individual attribute values are stored inline if they're small enough
(currently about 80 bytes, this can be changed though), and otherwise are
expanded to a btree. The theoretical limit to the size of an individual
attribute is about the same as an inode, though the kernel's upper bound on
the size of an attributes data is far smaller.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Make high level btree extend code generic
Tao Ma [Mon, 18 Aug 2008 09:38:45 +0000 (17:38 +0800)] 
ocfs2: Make high level btree extend code generic

Factor out the non-inode specifics of ocfs2_do_extend_allocation() into a more generic
function, ocfs2_do_cluster_allocation(). ocfs2_do_extend_allocation calls
ocfs2_do_cluster_allocation() now, but the latter can be used for other
btree types as well.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Abstract ocfs2_extent_tree in b-tree operations.
Tao Ma [Mon, 18 Aug 2008 09:38:44 +0000 (17:38 +0800)] 
ocfs2: Abstract ocfs2_extent_tree in b-tree operations.

In the old extent tree operation, we take the hypothesis that we
are using the ocfs2_extent_list in ocfs2_dinode as the tree root.
As xattr will also use ocfs2_extent_list to store large value
for a xattr entry, we refactor the tree operation so that xattr
can use it directly.

The refactoring includes 4 steps:
1. Abstract set/get of last_eb_blk and update_clusters since they may
   be stored in different location for dinode and xattr.
2. Add a new structure named ocfs2_extent_tree to indicate the
   extent tree the operation will work on.
3. Remove all the use of fe_bh and di, use root_bh and root_el in
   extent tree instead. So now all the fe_bh is replaced with
   et->root_bh, el with root_el accordingly.
4. Make ocfs2_lock_allocators generic. Now it is limited to be only used
   in file extend allocation. But the whole function is useful when we want
   to store large EAs.

Note: This patch doesn't touch ocfs2_commit_truncate() since it is not used
for anything other than truncate inode data btrees.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Use ocfs2_extent_list instead of ocfs2_dinode.
Tao Ma [Mon, 18 Aug 2008 09:38:43 +0000 (17:38 +0800)] 
ocfs2: Use ocfs2_extent_list instead of ocfs2_dinode.

ocfs2_extend_meta_needed(), ocfs2_calc_extend_credits() and
ocfs2_reserve_new_metadata() are all useful for extent tree operations. But
they are all limited to an inode btree because they use a struct
ocfs2_dinode parameter. Change their parameter to struct ocfs2_extent_list
(the part of an ocfs2_dinode they actually use) so that the xattr btree code
can use these functions.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Modify ocfs2_num_free_extents for future xattr usage.
Tao Ma [Mon, 18 Aug 2008 09:38:42 +0000 (17:38 +0800)] 
ocfs2: Modify ocfs2_num_free_extents for future xattr usage.

ocfs2_num_free_extents() is used to find the number of free extent records
in an inode btree. Hence, it takes an "ocfs2_dinode" parameter. We want to
use this for extended attribute trees in the future, so genericize the
interface the take a buffer head. A future patch will allow that buffer_head
to contain any structure rooting an ocfs2 btree.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: track local alloc state via debugfs
Mark Fasheh [Wed, 30 Jul 2008 01:29:18 +0000 (18:29 -0700)] 
ocfs2: track local alloc state via debugfs

A per-mount debugfs file, "local_alloc" is created which when read will
expose live state of the nodes local alloc file. Performance impact is
minimal, only a bit of memory overhead per mount point. Still, the code is
hidden behind CONFIG_OCFS2_FS_STATS. This feature will help us debug
local alloc performance problems on a live system.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: throttle back local alloc when low on disk space
Mark Fasheh [Tue, 29 Jul 2008 01:02:53 +0000 (18:02 -0700)] 
ocfs2: throttle back local alloc when low on disk space

Ocfs2's local allocator disables itself for the duration of a mount point
when it has trouble allocating a large enough area from the primary bitmap.
That can cause performance problems, especially for disks which were only
temporarily full or fragmented. This patch allows for the allocator to
shrink it's window first, before being disabled. Later, it can also be
re-enabled so that any performance drop is minimized.

To do this, we allow the value of osb->local_alloc_bits to be shrunk when
needed. The default value is recorded in a mostly read-only variable so that
we can re-initialize when required.

Locking had to be updated so that we could protect changes to
local_alloc_bits. Mostly this involves protecting various local alloc values
with the osb spinlock. A new state is also added, OCFS2_LA_THROTTLED, which
is used when the local allocator is has shrunk, but is not disabled. If the
available space dips below 1 megabyte, the local alloc file is disabled. In
either case, local alloc is re-enabled 30 seconds after the event, or when
an appropriate amount of bits is seen in the primary bitmap.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Track local alloc bits internally
Mark Fasheh [Mon, 28 Jul 2008 21:55:20 +0000 (14:55 -0700)] 
ocfs2: Track local alloc bits internally

Do this instead of tracking absolute local alloc size. This avoids
needless re-calculatiion of bits from bytes in localalloc.c. Additionally,
the value is now in a more natural unit for internal file system bitmap
work.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: POSIX file locks support
Mark Fasheh [Mon, 21 Jul 2008 21:29:16 +0000 (14:29 -0700)] 
ocfs2: POSIX file locks support

This is actually pretty easy since fs/dlm already handles the bulk of the
work. The Ocfs2 userspace cluster stack module already uses fs/dlm as the
underlying lock manager, so I only had to add the right calls.

Cluster-aware POSIX locks ("plocks") can be turned off by the same means at
UNIX locks - mount with 'noflocks', or create a local-only Ocfs2 volume.
Internally, the file system uses two sets of file_operations, depending on
whether cluster aware plocks is required. This turns out to be easier than
implementing local-only versions of ->lock.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agovfs: Use const for kernel parser table
Steven Whitehouse [Mon, 13 Oct 2008 09:46:57 +0000 (10:46 +0100)] 
vfs: Use const for kernel parser table

This is a much better version of a previous patch to make the parser
tables constant. Rather than changing the typedef, we put the "const" in
all the various places where its required, allowing the __initconst
exception for nfsroot which was the cause of the previous trouble.

This was posted for review some time ago and I believe its been in -mm
since then.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Alexander Viro <aviro@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cooloney...
Linus Torvalds [Mon, 13 Oct 2008 17:08:43 +0000 (10:08 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/cooloney/blackfin-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cooloney/blackfin-2.6: (60 commits)
  Blackfin arch: make sure we include the fix for SPORT hysteresis when reprogramming clocks
  Blackfin arch: Fix bogus str_ident check in gpio code
  Blackfin arch: AD7879 Touchscreen driver
  Blackfin arch: introducing bfin_addr_dcachable
  Blackfin arch: fix a typo in comments
  Blackfin arch: Remove useless head file
  Blackfin arch: make sure L2 start and length are always defined (fixes building on BF542)
  Blackfin arch: use the Blackfin on-chip ROM to do software reset when possible
  Blackfin arch: update anomaly headers to match the latest sheet
  Blackfin arch: bfin_reset() is an internal reboot function ... everyone should go through machine_restart()
  Blackfin arch: print out error/warning if you are running on the incorrect CPU type
  Blackfin arch: remove non-bf54x ifdef logic since this file is only compiled on bf54x parts
  Blackfin arch: update board defconfigs
  Blackfin arch: Add optional verbose debug
  Blackfin arch: emulate a TTY over the EMUDAT/JTAG interface
  Blackfin arch: have is_user_addr_valid() check for overflows (like when address is -1)
  Blackfin arch: ptrace - fix off-by-one check on end of memory regions
  Blackfin arch: Enable framebuffer support for the BF526-EZkit TFT LCD display
  Blackfin arch: flash memory map and dm9000 resources updating
  Blackfin arch: early prink code still use uart core console functions to parse and set configure option string
  ...

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Mon, 13 Oct 2008 17:08:08 +0000 (10:08 -0700)] 
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  net/mac80211/rx.c: fix build error
  acpi: Make ACPI_TOSHIBA depend on INPUT.
  net/bfin_mac.c MDIO namespace fixes
  jme: remove unused #include <version.h>
  netfilter: remove unused #include <version.h>
  net: Fix off-by-one in skb_dma_map
  smc911x: Add support for LAN921{5,7,8} chips from SMSC
  qlge: remove duplicated #include
  wireless: remove duplicated #include
  net/au1000_eth.c MDIO namespace fixes
  net/tc35815.c: fix compilation
  sky2: Fix WOL regression
  r8169: NULL pointer dereference on r8169 load

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6
Linus Torvalds [Mon, 13 Oct 2008 17:07:53 +0000 (10:07 -0700)] 
Merge git://git./linux/kernel/git/davem/sparc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
  chmc: Mark %ver register inline asm with __volatile__
  sparc64: Add missing notify_cpu_starting() call.
  sparc32: fix build errors

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Mon, 13 Oct 2008 17:06:58 +0000 (10:06 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6: (33 commits)
  ALSA: ASoC codec: remove unused #include <version.h>
  ALSA: ASoC: update email address for Liam Girdwood
  ALSA: hda: corrected invalid mixer values
  ALSA: hda: add mixers for analog mixer on 92hd75xx codecs
  ALSA: ASoC: Add destination and source port for DMA on OMAP1
  ALSA: ASoC: Drop device registration from GTA01 lm4857 driver
  ALSA: ASoC: Fix build of GTA01 audio driver
  ALSA: ASoC: Add widgets before setting endpoints on GTA01
  ALSA: ASoC: Fix inverted input PGA mute bits in WM8903
  ALSA: ASoC: OMAP: Set DMA stream name at runtime in McBSP DAI driver
  ALSA: ASoC: OMAP: Add support for OMAP2430 and OMAP34xx in McBSP DAI driver
  ALSA: ASoC: OMAP: Add multilink support to McBSP DAI driver
  ALSA: ASoC: Make TLV320AIC26 user-visible
  ALSA: ASoC - clean up Kconfig for TLV320AIC2
  ALSA: ASoC: Make WM8510 microphone input a DAPM mixer
  ALSA: ASoC: Implement WM8510 bias level control
  ALSA: ASoC: Remove unused AUDIO_NAME define from codec drivers
  ALSA: ASoC: tlv320aic3x: Use uniform tlv320aic naming
  ALSA: ASoC: Add WM8510 SPI support
  ALSA: ASoC: Add WM8753 SPI support
  ...

16 years agoMerge branch 'proc' of git://git.kernel.org/pub/scm/linux/kernel/git/adobriyan/proc
Linus Torvalds [Mon, 13 Oct 2008 17:04:04 +0000 (10:04 -0700)] 
Merge branch 'proc' of git://git./linux/kernel/git/adobriyan/proc

* 'proc' of git://git.kernel.org/pub/scm/linux/kernel/git/adobriyan/proc:
  proc: remove kernel.maps_protect
  proc: remove now unneeded ADDBUF macro
  [PATCH] proc: show personality via /proc/pid/personality
  [PATCH] signal, procfs: some lock_task_sighand() users do not need rcu_read_lock()
  proc: move PROC_PAGE_MONITOR to fs/proc/Kconfig
  proc: make grab_header() static
  proc: remove unused get_dma_list()
  proc: remove dummy vmcore_open()
  proc: proc_sys_root tweak
  proc: fix return value of proc_reg_open() in "too late" case

Fixed up trivial conflict in removed file arch/sparc/include/asm/dma_32.h

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Mon, 13 Oct 2008 17:00:44 +0000 (10:00 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/jmorris/security-testing-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/security-testing-2.6: (24 commits)
  integrity: special fs magic
  As pointed out by Jonathan Corbet, the timer must be deleted before
  ERROR: code indent should use tabs where possible
  The tpm_dev_release function is only called for platform devices, not pnp
  Protect tpm_chip_list when transversing it.
  Renames num_open to is_open, as only one process can open the file at a time.
  Remove the BKL calls from the TPM driver, which were added in the overall
  netlabel: Add configuration support for local labeling
  cipso: Add support for native local labeling and fixup mapping names
  netlabel: Changes to the NetLabel security attributes to allow LSMs to pass full contexts
  selinux: Cache NetLabel secattrs in the socket's security struct
  selinux: Set socket NetLabel based on connection endpoint
  netlabel: Add functionality to set the security attributes of a packet
  netlabel: Add network address selectors to the NetLabel/LSM domain mapping
  netlabel: Add a generic way to create ordered linked lists of network addrs
  netlabel: Replace protocol/NetLabel linking with refrerence counts
  smack: Fix missing calls to netlbl_skbuff_err()
  selinux: Fix missing calls to netlbl_skbuff_err()
  selinux: Fix a problem in security_netlbl_sid_to_secattr()
  selinux: Better local/forward check in selinux_ip_postroute()
  ...

16 years agoMerge git://git.infradead.org/users/dwmw2/random-2.6
Linus Torvalds [Mon, 13 Oct 2008 16:59:14 +0000 (09:59 -0700)] 
Merge git://git.infradead.org/users/dwmw2/random-2.6

* git://git.infradead.org/users/dwmw2/random-2.6:
  Fix autoloading of MacBook Pro backlight driver.
  Automatic MODULE_ALIAS() for DMI match tables.
  Remove asm/a.out.h files for all architectures without a.out support.
  Introduce HAVE_AOUT symbol to remove hard-coded arch list for BINFMT_AOUT
  Remove redundant CONFIG_ARCH_SUPPORTS_AOUT
  S390: Update comments about why we don't use <asm-generic/statfs.h>
  SPARC: Use <asm-generic/statfs.h>
  PowerPC: Use <asm-generic/statfs.h>
  PARISC: Use <asm-generic/statfs.h>
  x86_64: Use <asm-generic/statfs.h>
  IA64: Use <asm-generic/statfs.h>
  ARM: Use <asm-generic/statfs.h>
  Make <asm-generic/statfs.h> suitable for 64-bit platforms.
  Define and use PCI_DEVICE_ID_MARVELL_88ALP01_CCIC for CAFÉ camera driver
  [MTD] [NAND] Define and use PCI_DEVICE_ID_MARVELL_88ALP01_NAND for CAFÉ
  Use PCI_DEVICE_ID_88ALP01 for CAFÉ chip, rather than PCI_DEVICE_ID_CAFE.
  EFS: Don't set f_fsid in statfs().

16 years agoMerge phase #5 (misc) of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux...
Linus Torvalds [Mon, 13 Oct 2008 16:54:35 +0000 (09:54 -0700)] 
Merge phase #5 (misc) of git://git./linux/kernel/git/tip/linux-2.6-tip

Merges oprofile, timers/hpet, x86/traps, x86/time, and x86/core misc items.

* 'x86-core-v4-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (132 commits)
  x86: change early_ioremap to use slots instead of nesting
  x86: adjust dependencies for CONFIG_X86_CMOV
  dumpstack: x86: various small unification steps, fix
  x86: remove additional_cpus
  x86: remove additional_cpus configurability
  x86: improve UP kernel when CPU-hotplug and SMP is enabled
  dumpstack: x86: various small unification steps
  dumpstack: i386: make kstack= an early boot-param and add oops=panic
  dumpstack: x86: use log_lvl and unify trace formatting
  dumptrace: x86: consistently include loglevel, print stack switch
  dumpstack: x86: add "end" parameter to valid_stack_ptr and print_context_stack
  dumpstack: x86: make printk_address equal
  dumpstack: x86: move die_nmi to dumpstack_32.c
  traps: x86: finalize unification of traps.c
  traps: x86: make traps_32.c and traps_64.c equal
  traps: x86: various noop-changes preparing for unification of traps_xx.c
  traps: x86_64: use task_pid_nr(tsk) instead of tsk->pid in do_general_protection
  traps: i386: expand clear_mem_error and remove from mach_traps.h
  traps: x86_64: make io_check_error equal to the one on i386
  traps: i386: use preempt_conditional_sti/cli in do_int3
  ...

16 years agotty: rename the remaining oddly named n_tty functions
Alan Cox [Mon, 13 Oct 2008 09:46:24 +0000 (10:46 +0100)] 
tty: rename the remaining oddly named n_tty functions

Original idea for this from a patch by Rodolfo Giometti which merges various
bits of PPS support

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agofs3270: Correct error returns
Alan Cox [Mon, 13 Oct 2008 09:46:09 +0000 (10:46 +0100)] 
fs3270: Correct error returns

Drop the kernel lock further and also correct cases where we set rc to an
error code, and then return 0

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agofs3270: remove extra locks
Alan Cox [Mon, 13 Oct 2008 09:45:52 +0000 (10:45 +0100)] 
fs3270: remove extra locks

get_current_tty now does internal locking and returns a referenced object,
thus our use of tty_mutex here can go away.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: tty_io.c shadows sparse fix
Jason Wessel [Mon, 13 Oct 2008 09:45:36 +0000 (10:45 +0100)] 
tty: tty_io.c shadows sparse fix

drivers/char/tty_io.c:1413:17: warning: symbol 'buf' shadows an earlier one
drivers/char/tty_io.c:1379:20: originally declared here

Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoserial: fix device name reporting when minor space is shared between drivers
David S. Miller [Mon, 13 Oct 2008 09:45:26 +0000 (10:45 +0100)] 
serial: fix device name reporting when minor space is shared between drivers

The multiple drivers share the minor space occupied by a particular major
number, the actual index within the device name's space is indicated by
the tty_driver->name_base + uart_port->line

Another usable formula is (uart_driver->minor - MINOR_BASE) + port->line

Use those to print the device names properly in such situations in
serial_core.c and 8250.c

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoapplicom: Fix an unchecked user ioctl range and an error return
Alan Cox [Mon, 13 Oct 2008 09:45:17 +0000 (10:45 +0100)] 
applicom: Fix an unchecked user ioctl range and an error return

Closes bug #11408 by checking the card index range for command 0
Fixes the ioctl to return ENOTTY which is correct for unknown ioctls

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Minor tidyups and document fixes for n_tty
Alan Cox [Mon, 13 Oct 2008 09:45:06 +0000 (10:45 +0100)] 
tty: Minor tidyups and document fixes for n_tty

Remove/fix some bogus NULL checks, comment some locking etc

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Remove lots of NULL checks
Alan Cox [Mon, 13 Oct 2008 09:44:57 +0000 (10:44 +0100)] 
tty: Remove lots of NULL checks

Many tty drivers contain 'can't happen' checks against NULL pointers passed
in by the tty layer. These have never been possible to occur. Even more
importantly if they ever do occur we want to know as it would be a serious
bug.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: fix up gigaset a bit
Alan Cox [Mon, 13 Oct 2008 09:44:43 +0000 (10:44 +0100)] 
tty: fix up gigaset a bit

Stephen's fixes reminded me that gigaset is still rather broken so fix it up
a bit

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Fallout from tty-move-canon-specials
Stephen Rothwell [Mon, 13 Oct 2008 09:44:33 +0000 (10:44 +0100)] 
tty: Fallout from tty-move-canon-specials

Today's linux-next build (x86_64 allmodconfig) failed like this:

/drivers/char/tty_ioctl.c: In function 'change_termios':
drivers/isdn/capi/capi.c:1234: error: implicit declaration of function 'n_tty_ioctl'
drivers/isdn/gigaset/ser-gigaset.c: In function 'gigaset_tty_ioctl':
drivers/isdn/gigaset/ser-gigaset.c:648: error: implicit declaration of function 'n_tty_ioctl'

Introduced by commit 686b5e4aea05a80e370dc931b7f4a8d03c80da54
("tty-move-canon-specials").  I added the following patch (which may not
be correct).

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: some ICANON magic is in the wrong places
Alan Cox [Mon, 13 Oct 2008 09:44:17 +0000 (10:44 +0100)] 
tty: some ICANON magic is in the wrong places

Move the set up on ldisc change into the ldisc
Move the INQ/OUTQ cases into the driver not in shared ioctl code where it
gives bogus answers for other ldisc values

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: simplify ktermios allocation
Alan Cox [Mon, 13 Oct 2008 09:44:08 +0000 (10:44 +0100)] 
tty: simplify ktermios allocation

Copy the simplification from the pty unix98 special case to the generic one.
This allows us to kill off driver->termios_locked entirely which is nice. We
have to whack bits of the cris driver as it meddles in places it shouldn't
providing its own arrays that were never used anyway.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agopty: simplify unix98 allocation
Alan Cox [Mon, 13 Oct 2008 09:43:58 +0000 (10:43 +0100)] 
pty: simplify unix98 allocation

We need both termios and termios_locked so allocate them as one

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agopty: Fix allocation failure double free
Alan Cox [Mon, 13 Oct 2008 09:43:48 +0000 (10:43 +0100)] 
pty: Fix allocation failure double free

The updating and moving around of the pty code added a bug where both the
helper and caller free the main tty struct (the pty driver must free the
o_tty pair itself however).

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agopty: Coding style and polish
Alan Cox [Mon, 13 Oct 2008 09:43:38 +0000 (10:43 +0100)] 
pty: Coding style and polish

We've done the heavy lifting now its time to mop up a bit

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoSimplify devpts_pty_kill
Sukadev Bhattiprolu [Mon, 13 Oct 2008 09:43:27 +0000 (10:43 +0100)] 
Simplify devpts_pty_kill

When creating a new pty, save the pty's inode in the tty->driver_data.
Use this inode in pty_kill() to identify the devpts instance. Since
we now have the inode for the pty, we can skip get_node() lookup and
remove the unused get_node().

TODO:
- check if the mutex_lock is needed in pty_kill().

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoSimplify devpts_pty_new()
Sukadev Bhattiprolu [Mon, 13 Oct 2008 09:43:18 +0000 (10:43 +0100)] 
Simplify devpts_pty_new()

devpts_pty_new() is called when setting up a new pty and would not
will not have an existing dentry or inode for the pty. So don't bother
looking for an existing dentry - just create a new one.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoSimplify devpts_get_tty()
Sukadev Bhattiprolu [Mon, 13 Oct 2008 09:43:08 +0000 (10:43 +0100)] 
Simplify devpts_get_tty()

As pointed out by H. Peter Anvin, since the inode for the pty is known,
we don't need to look it up.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoAdd an instance parameter devpts interfaces
Sukadev Bhattiprolu [Mon, 13 Oct 2008 09:42:59 +0000 (10:42 +0100)] 
Add an instance parameter devpts interfaces

Pass-in 'inode' or 'tty' parameter to devpts interfaces.  With multiple
devpts instances, these parameters will be used in subsequent patches
to identify the instance of devpts mounted. The parameters also help
simplify devpts implementation.

Changelog[v3]:
- minor changes due to merge with ttydev updates
- rename parameters to emphasize they are ptmx or pts inodes
- pass-in tty_struct * to devpts_pty_kill() (this will help
  cleanup the get_node() call in a subsequent patch)

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMove tty lookup/reopen to caller
Sukadev Bhattiprolu [Mon, 13 Oct 2008 09:42:49 +0000 (10:42 +0100)] 
Move tty lookup/reopen to caller

Move tty_driver_lookup_tty() and tty_reopen() from tty_init_dev()
into tty_open() (one of the two callers of tty_init_dev()).  These
calls are not really required in ptmx_open(), the other caller,
since ptmx_open() would be setting up a new tty.

Changelog[v2]:
- remove the lookup and reopen calls from ptmx_open
- merge with recent changes to ttydev tree

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: extract the pty init time special cases
Alan Cox [Mon, 13 Oct 2008 09:42:39 +0000 (10:42 +0100)] 
tty: extract the pty init time special cases

The majority of the remaining init_dev code is pty special cases. We
refactor this code into the driver->install method.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Finish fixing up the init_dev interface to use ERR_PTR
Alan Cox [Mon, 13 Oct 2008 09:42:29 +0000 (10:42 +0100)] 
tty: Finish fixing up the init_dev interface to use ERR_PTR

Original suggestion and proposal from Sukadev Bhattiprolu.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: More driver operations
Alan Cox [Mon, 13 Oct 2008 09:42:19 +0000 (10:42 +0100)] 
tty: More driver operations

We have the lookup operation abstracted which is nice for pty cleanup but
we really want to abstract the add/remove entries as well so that we can
pull the pty code out of the tty core and create a clear defined interface
for the tty driver table.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: kref the tty driver object
Alan Cox [Mon, 13 Oct 2008 09:42:09 +0000 (10:42 +0100)] 
tty: kref the tty driver object

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Clean up the tty_init_dev changes further
Alan Cox [Mon, 13 Oct 2008 09:42:00 +0000 (10:42 +0100)] 
tty: Clean up the tty_init_dev changes further

Fix up the naming, style and extract some bits of code into the driver
specific code

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Move parts of tty_init_dev into new functions
Sukadev Bhattiprolu [Mon, 13 Oct 2008 09:41:51 +0000 (10:41 +0100)] 
tty: Move parts of tty_init_dev into new functions

Move the 'find-tty' and 'fast-track-open' parts of init_dev() to
separate functions.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Remove more special casing and out of place code
Alan Cox [Mon, 13 Oct 2008 09:41:42 +0000 (10:41 +0100)] 
tty: Remove more special casing and out of place code

Carry on pushing code out of tty_io when it belongs to other drivers. I'm
not 100% happy with some of this and it will be worth revisiting some of the
exports later when the restructuring work is done.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: shutdown method
Alan Cox [Mon, 13 Oct 2008 09:41:30 +0000 (10:41 +0100)] 
tty: shutdown method

Right now there are various drivers that try to use tty->count to know when
they get the final close. Aristeau Rozanski showed while debugging the vt
sysfs race that this isn't entirely safe.

Instead of driver side tricks to work around this introduce a shutdown which
is called when the tty is being destructed. This also means that the shutdown
method is tied into the refcounting.

Use this to rework the console close/sysfs logic.

Remove lots of special case code from the tty core code. The pty code can now
have a shutdown() method that replaces the special case hackery in the tree
free up paths.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agovt: remove bogus lock dropping
Alan Cox [Mon, 13 Oct 2008 09:41:16 +0000 (10:41 +0100)] 
vt: remove bogus lock dropping

For hysterical raisins the vt layer drops and retakes locks in the write
method. This is a left over from the days when user/kernel data was passed
directly to the tty not pre-buffered.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agopty: If the administrator creates a device for a ptmx slave we should not error
Alan Cox [Mon, 13 Oct 2008 09:41:03 +0000 (10:41 +0100)] 
pty: If the administrator creates a device for a ptmx slave we should not error

The open path for ptmx slaves is via the ptmx device. Opening them any
other way is not allowed. Vegard Nossum found that previously this was not
the case and mknod foo c 128 42; cat foo would produce nasty diagnostics

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Fix abusers of current->sighand->tty
Alan Cox [Mon, 13 Oct 2008 09:40:53 +0000 (10:40 +0100)] 
tty: Fix abusers of current->sighand->tty

Various people outside the tty layer still stick their noses in behind the
scenes. We need to make sure they also obey the locking and referencing rules.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Redo current tty locking
Alan Cox [Mon, 13 Oct 2008 09:40:43 +0000 (10:40 +0100)] 
tty: Redo current tty locking

Currently it is sometimes locked by the tty mutex and sometimes by the
sighand lock. The latter is in fact correct and now we can hand back referenced
objects we can fix this up without problems around sleeping functions.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: the vhangup syscall is racy
Alan Cox [Mon, 13 Oct 2008 09:40:30 +0000 (10:40 +0100)] 
tty: the vhangup syscall is racy

We now have the infrastructure to sort this out but rather than teaching
the syscall tty lock rules we move the hard work into a tty helper

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomxser: Switch to kref tty
Alan Cox [Mon, 13 Oct 2008 09:40:19 +0000 (10:40 +0100)] 
mxser: Switch to kref tty

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agostallion: Use krefs
Alan Cox [Mon, 13 Oct 2008 09:40:07 +0000 (10:40 +0100)] 
stallion: Use krefs

Use tty_port_init and krefs in the stallion drivers to protect us from devices
going away underneath us. As with the other drives some rearranging is done to
pass the tty structure down properly on the user side.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: kref usage for isicom and moxa
Alan Cox [Mon, 13 Oct 2008 09:39:58 +0000 (10:39 +0100)] 
tty: kref usage for isicom and moxa

Rather than blindly keep taking krefs we reorder the code in a few places
to pass the tty down to the right place (which is important as from the user
side it is not the case that tty == port->tty in all situations). For the irq
and related paths use the krefs to stop the tty being freed under us.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: usb-serial krefs
Alan Cox [Mon, 13 Oct 2008 09:39:46 +0000 (10:39 +0100)] 
tty: usb-serial krefs

Use kref in the USB serial drivers so that we don't free tty structures
from under the URB receive handlers as has historically been the case if
you were unlucky. This also gives us a framework for general tty drivers to
use tty_port objects and refcount.

Contains two err->dev_err changes merged together to fix clashes in the
-next tree.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Move tty_write_message out of kernel/printk
Alan Cox [Mon, 13 Oct 2008 09:39:23 +0000 (10:39 +0100)] 
tty: Move tty_write_message out of kernel/printk

This is pure tty code so put it in the tty layer where it can be with the
locking relevant material it uses

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Make get_current_tty use a kref
Alan Cox [Mon, 13 Oct 2008 09:39:13 +0000 (10:39 +0100)] 
tty: Make get_current_tty use a kref

We now return a kref covered tty reference. That ensures the tty structure
doesn't go away when you have a return from get_current_tty. This is not
enough to protect you from most of the resources being freed behind your
back - yet.

[Updated to include fixes for SELinux problems found by Andrew Morton and
 an s390 leak found while debugging the former]

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: compare the tty winsize
Alan Cox [Mon, 13 Oct 2008 09:39:01 +0000 (10:39 +0100)] 
tty: compare the tty winsize

We always use the real tty one for stuff so the pty one should not be
compared. As we propagate window changes to both it doesn't currently
matter but will when we tidy up the pty termios logic a bit more

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Termios locking - sort out real_tty confusions and lock reads
Alan Cox [Mon, 13 Oct 2008 09:38:46 +0000 (10:38 +0100)] 
tty: Termios locking - sort out real_tty confusions and lock reads

This moves us towards sanity and should mean our termios locking is now
complete and comprehensive.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Add termiox
Alan Cox [Mon, 13 Oct 2008 09:38:18 +0000 (10:38 +0100)] 
tty: Add termiox

We need a way to describe the various additional modes and flow control
features that random weird hardware shows up and software such as wine
wants to emulate as Windows supports them.

TCGETX/TCSETX and the termiox ioctl are a SYS5 extension that we might as
well adopt. This patches adds the structures and the basic ioctl interfaces
when the TCGETX etc defines are added for an architecture. Drivers wishing
to use this stuff need to add new methods.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: ipw need reworking
Alan Cox [Mon, 13 Oct 2008 09:38:07 +0000 (10:38 +0100)] 
tty: ipw need reworking

This came in via another tree and unfortunately is rather broken on
the tty side. Comment the apparent locking problems for someone who knows
the driver to look at.

Fix the termios and other ioctl handling. The driver was calling the wrong
methods for what it wanted to do but the right ones existed so its a simple
fix up.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Cris has a nice RS485 ioctl so we should steal it
Alan Cox [Mon, 13 Oct 2008 09:37:48 +0000 (10:37 +0100)] 
tty: Cris has a nice RS485 ioctl so we should steal it

JP Tosoni observed:

"About a RS485 ioctl: could you consider the attached files which are
 already in the Linux kernel (in include/asm-cris).  They define a
 TIOCSERSETRS485 (ioctl.h), and the data structure (rs485.h)
 with allows to specify timings. Sounds just like what we want ?"

and he's right: sort of. Rework the structure to use flag bits and make the
time delay a fixed sized field so we don't get 32/64bit problems. Add the ioctls
to x86 so that people know what to add to their platform of choice.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: use krefs to protect driver module counts
Alan Cox [Mon, 13 Oct 2008 09:37:36 +0000 (10:37 +0100)] 
tty: use krefs to protect driver module counts

The tty layer keeps driver module counts that are used so the driver knows
when it can be unloaded. For obvious reasons we want to tie that to the
refcounting properly.

At this point the driver side itself isn't refcounted nicely but we can do
that later and kref the drivers.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Add a kref count
Alan Cox [Mon, 13 Oct 2008 09:37:26 +0000 (10:37 +0100)] 
tty: Add a kref count

Introduce a kref to the tty structure and use it to protect the tty->signal
tty references. For now we don't introduce it for anything else.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agopps: Reserve a line discipline number for PPS
Alan Cox [Mon, 13 Oct 2008 09:37:17 +0000 (10:37 +0100)] 
pps: Reserve a line discipline number for PPS

Add a new line discipline for "pulse per second" devices connected to
a serial port.

Signed-off-by: Rodolfo Giometti <giometti@linux.it>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: Split tty_port into its own file
Alan Cox [Mon, 13 Oct 2008 09:37:07 +0000 (10:37 +0100)] 
tty: Split tty_port into its own file

Not much in it yet but this will grow a lot

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: split the buffering from tty_io
Alan Cox [Mon, 13 Oct 2008 09:36:58 +0000 (10:36 +0100)] 
tty: split the buffering from tty_io

The two are basically independent chunks of code so lets split them up for
readability and sanity. It also makes the API boundaries much clearer.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agouml: small cleanups and note bugs to be dealt with by uml authors...
Alan Cox [Mon, 13 Oct 2008 09:36:49 +0000 (10:36 +0100)] 
uml: small cleanups and note bugs to be dealt with by uml authors...

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agotty: move tioclinux from a special case
Alan Cox [Mon, 13 Oct 2008 09:36:40 +0000 (10:36 +0100)] 
tty: move tioclinux from a special case

Right now we have ifdefs and hooks in the core ioctl handler for TIOCLINUX
and then test if its a console. This is brain dead. Instead call the
tioclinux helper from the relevant driver ioctl methods.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoserial: allow 8250 to be used on sparc
David Miller [Mon, 13 Oct 2008 09:36:31 +0000 (10:36 +0100)] 
serial: allow 8250 to be used on sparc

This requires three changes:

1) Remove !SPARC restriction in Kconfig.

2) Move Sparc specific serial drivers before 8250, so that serial
   console devices don't change names on us, even if 8250 finds
   devices.

3) Since the Sparc specific serial drivers try to use the
   same major/minor device namespace as 8250, some coordination
   is necessary.  Use the sunserial_*() layer routines to allocate
   minor number space within TTY_MAJOR when CONFIG_SPARC.

   This has no effect on other platforms.

Thanks to Josip Rodin for bringing up this issue and testing
plus debugging various revisions of this patch.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years ago8250: remove a few inlines of dubious value
Will Newton [Mon, 13 Oct 2008 09:36:21 +0000 (10:36 +0100)] 
8250: remove a few inlines of dubious value

Remove some inlines from various functions that are called once, are too
big to inline, or are called only from slow path code.  This saves around
300 bytes of code for me.

Signed-off-by: Will Newton <will.newton@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoserial_8250: pci_enable_device fail is not fully handled
Alan Cox [Mon, 13 Oct 2008 09:36:11 +0000 (10:36 +0100)] 
serial_8250: pci_enable_device fail is not fully handled

<rmk> talking about leaks - I noticed that the 'check return of
pci_enable_dev()' in the 8250 pci resume function finally made it in
despite my objections against it (causing stuff in higher levels to
leak).

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoftdi: A few errors are err() that should be debug which causes much spewage
Alan Cox [Mon, 13 Oct 2008 09:36:00 +0000 (10:36 +0100)] 
ftdi: A few errors are err() that should be debug which causes much spewage

Fixes #10783

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agousb: fix pl2303 initialization
Jason Wessel [Mon, 13 Oct 2008 09:35:51 +0000 (10:35 +0100)] 
usb: fix pl2303 initialization

This patch removes the private check for the termios_initialized for
the pl2303 usb driver.  It forced the baud to 9600 on the first call
to pl2303_set_termios()

Based on the tty changes in the 2.6.27 kernel, the termios passed to
the *_set_termios functions is always populated the first time.

This means there is no need to privately initialize the settings the
first time, and doing so will not allow the use of the kernel
parameter "console=ttyUSB0,115200" as an example.

Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoserial-make-uart_ports-ioport-unsigned-long-fix
Andrew Morton [Mon, 13 Oct 2008 09:35:42 +0000 (10:35 +0100)] 
serial-make-uart_ports-ioport-unsigned-long-fix

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agonozomi: Fix close on error
Alan Cox [Mon, 13 Oct 2008 09:35:33 +0000 (10:35 +0100)] 
nozomi: Fix close on error

Nozomi assumes the close method isn't called if open errors. The tty layer
is different to other drives in this respect however. Pointed out by Denis J
Barrow.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoserial: Make uart_port's ioport "unsigned long".
David Miller [Mon, 13 Oct 2008 09:35:23 +0000 (10:35 +0100)] 
serial: Make uart_port's ioport "unsigned long".

Otherwise the top 32-bits of the resource value get chopped
off on 64-bit systems, and the resulting I/O accesses go to
random places.

Thanks to testing and debugging by Josip Rodin, which helped
track this down.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoaudit: Handle embedded NUL in TTY input auditing
Miloslav Trmac [Mon, 13 Oct 2008 09:35:15 +0000 (10:35 +0100)] 
audit: Handle embedded NUL in TTY input auditing

Data read from a TTY can contain an embedded NUL byte (e.g. after
pressing Ctrl-2, or sent to a PTY).  After the previous patch, the data
would be logged only up to the first NUL.

This patch modifies the AUDIT_TTY record to always use the hexadecimal
format, which does not terminate at the first NUL byte.  The vast
majority of recorded TTY input data will contain either ' ' or '\n', so
the hexadecimal format would have been used anyway.

Signed-off-by: Miloslav Trmac <mitr@redhat.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoip2: avoid add_timer with pending timer
Akinobu Mita [Mon, 13 Oct 2008 09:35:05 +0000 (10:35 +0100)] 
ip2: avoid add_timer with pending timer

add_timer() is not supposed to be called when the timer is pending.
ip2 driver attempts to avoid that condition by setting and resetting
a flag (TimerOn) in timer function. But there is some gap between
add_timer() and setting TimerOn.

This patch fix this problem by using mod_timer() and remove TimerOn
which has been unnecessary by this change.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoip2: init/deinit cleanup
Jiri Slaby [Mon, 13 Oct 2008 09:34:56 +0000 (10:34 +0100)] 
ip2: init/deinit cleanup

Cleanup of module_init/exit:
- mostly whitespace
- remove empty functions
- replace c++ comments
- remove useless prints (module loaded, unloaded)
- mark the calls as __exit and __init
- use break; and return; to save some indent levels after it
- note resource leakage

It's still mess, but now it's readable.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoip2: fix sparse warnings
Jiri Slaby [Mon, 13 Oct 2008 09:34:45 +0000 (10:34 +0100)] 
ip2: fix sparse warnings

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoip2: cleanup globals
Jiri Slaby [Mon, 13 Oct 2008 09:34:36 +0000 (10:34 +0100)] 
ip2: cleanup globals

- do not init .bss zeroed data to zero again (by memset or
  explicit assignment)
- use char [] instead of char * for string constants

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoChar: merge ip2main and ip2base
Jiri Slaby [Mon, 13 Oct 2008 09:34:27 +0000 (10:34 +0100)] 
Char: merge ip2main and ip2base

It's pretty useless to have one setup() function separated along with
module_init() which only calls a function from ip2main anyway. Get rid
of ip2base.

Remove also checks of always-true now.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoChar: sx, fix io unmapping
Jiri Slaby [Mon, 13 Oct 2008 09:34:18 +0000 (10:34 +0100)] 
Char: sx, fix io unmapping

board->base is increased for CF cards after mapping. Use board->base2
for unmapping the region, since it holds the original/correct address.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoChar: cyclades. remove bogus iomap
Jiri Slaby [Mon, 13 Oct 2008 09:34:09 +0000 (10:34 +0100)] 
Char: cyclades. remove bogus iomap

readl/writel are not expected to accept iomap return value. Replace
bogus mapping by standard ioremap.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoFix oti6858 debug level
Scott Ashcroft [Mon, 13 Oct 2008 09:34:00 +0000 (10:34 +0100)] 
Fix oti6858 debug level

For some reason the oti6858 driver undefines and redefines the dbg
macro. This makes it spew debugging messages at KERN_INFO instead of
KERN_DEBUG.

This patch removes the undef and define making the driver log like every
other USB serial driver.

Signed-off-by: Scott Ashcroft <scott.ashcroft@talk21.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoBlackfin Serial Driver: Fix bug - request UART2/3 peripheral mapped interrupts in...
Sonic Zhang [Mon, 13 Oct 2008 09:33:51 +0000 (10:33 +0100)] 
Blackfin Serial Driver: Fix bug - request UART2/3 peripheral mapped interrupts in PIO mode

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoBlackfin Serial Driver: Fix bug - ircp fails on sir over Blackfin UART
Graf Yang [Mon, 13 Oct 2008 09:33:42 +0000 (10:33 +0100)] 
Blackfin Serial Driver: Fix bug - ircp fails on sir over Blackfin UART

We now use the sir_dev/irtty_sir/uart/bfin_serial drivers framework
to monitor the TX status.

Signed-off-by: Graf Yang <graf.yang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoBlackfin Serial Driver: Fix bug - Don't call tx_stop in tx_transfer.
Sonic Zhang [Mon, 13 Oct 2008 09:33:33 +0000 (10:33 +0100)] 
Blackfin Serial Driver: Fix bug - Don't call tx_stop in tx_transfer.

Disable irq and return immediately.

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoBlackfin Serial Driver: Remove useless stop
Sonic Zhang [Mon, 13 Oct 2008 09:33:25 +0000 (10:33 +0100)] 
Blackfin Serial Driver: Remove useless stop

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoBlackfin Serial Driver: move common variables out of serial headers and into the...
Mike Frysinger [Mon, 13 Oct 2008 09:33:16 +0000 (10:33 +0100)] 
Blackfin Serial Driver: move common variables out of serial headers and into the serial driver

move common variables out of serial headers and into the serial driver and
rename "nr_ports" to "nr_active_ports" so as to easily differentiate
between BFIN_UART_NR_PORTS (the # of available) and nr_ports (the # of enabled)

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoBlackfin Serial Driver: trim trailing whitespace -- no functional changes
Mike Frysinger [Mon, 13 Oct 2008 09:33:06 +0000 (10:33 +0100)] 
Blackfin Serial Driver: trim trailing whitespace -- no functional changes

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoBlackfin Serial Driver: Fix bug - should suspend/resume/remove all uart ports.
Sonic Zhang [Mon, 13 Oct 2008 09:32:44 +0000 (10:32 +0100)] 
Blackfin Serial Driver: Fix bug - should suspend/resume/remove all uart ports.

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoBlackfin Serial Driver: use __initdata for data, not __init
Mike Frysinger [Mon, 13 Oct 2008 09:32:35 +0000 (10:32 +0100)] 
Blackfin Serial Driver: use __initdata for data, not __init

Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This page took 0.050949 seconds and 5 git commands to generate.