[deliverable/linux.git] / drivers / misc / cxl / context.c

/*
 * Copyright 2014 IBM Corp.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version
 * 2 of the License, or (at your option) any later version.
 */

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/bitmap.h>
#include <linux/sched.h>
#include <linux/pid.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/debugfs.h>
#include <linux/slab.h>
#include <linux/idr.h>
#include <asm/cputable.h>
#include <asm/current.h>
#include <asm/copro.h>

#include "cxl.h"

/*
 * Allocates space for a CXL context.
 */
struct cxl_context *cxl_context_alloc(void)
{
	return kzalloc(sizeof(struct cxl_context), GFP_KERNEL);
}

/*
 * Initialises a CXL context.
 */
int cxl_context_init(struct cxl_context *ctx, struct cxl_afu *afu, bool master,
		     struct address_space *mapping)
{
	int i;

	spin_lock_init(&ctx->sste_lock);
	ctx->afu = afu;
	ctx->master = master;
	ctx->pid = ctx->glpid = NULL; /* Set in start work ioctl */
	mutex_init(&ctx->mapping_lock);
	ctx->mapping = mapping;

	/*
	 * Allocate the segment table before we put it in the IDR so that we
	 * can always access it when dereferenced from IDR. For the same
	 * reason, the segment table is only destroyed after the context is
	 * removed from the IDR.  Access to this in the IOCTL is protected by
	 * Linux filesytem symantics (can't IOCTL until open is complete).
	 */
	i = cxl_alloc_sst(ctx);
	if (i)
		return i;

	INIT_WORK(&ctx->fault_work, cxl_handle_fault);

	init_waitqueue_head(&ctx->wq);
	spin_lock_init(&ctx->lock);

	ctx->irq_bitmap = NULL;
	ctx->pending_irq = false;
	ctx->pending_fault = false;
	ctx->pending_afu_err = false;

	INIT_LIST_HEAD(&ctx->irq_names);
	INIT_LIST_HEAD(&ctx->extra_irq_contexts);

	/*
	 * When we have to destroy all contexts in cxl_context_detach_all() we
	 * end up with afu_release_irqs() called from inside a
	 * idr_for_each_entry(). Hence we need to make sure that anything
	 * dereferenced from this IDR is ok before we allocate the IDR here.
	 * This clears out the IRQ ranges to ensure this.
	 */
	for (i = 0; i < CXL_IRQ_RANGES; i++)
		ctx->irqs.range[i] = 0;

	mutex_init(&ctx->status_mutex);

	ctx->status = OPENED;

	/*
	 * Allocating IDR! We better make sure everything's setup that
	 * dereferences from it.
	 */
	mutex_lock(&afu->contexts_lock);
	idr_preload(GFP_KERNEL);
	i = idr_alloc(&ctx->afu->contexts_idr, ctx,
		      ctx->afu->adapter->native->sl_ops->min_pe,
		      ctx->afu->num_procs, GFP_NOWAIT);
	idr_preload_end();
	mutex_unlock(&afu->contexts_lock);
	if (i < 0)
		return i;

	ctx->pe = i;
	if (cpu_has_feature(CPU_FTR_HVMODE)) {
		ctx->elem = &ctx->afu->native->spa[i];
		ctx->external_pe = ctx->pe;
	} else {
		ctx->external_pe = -1; /* assigned when attaching */
	}
	ctx->pe_inserted = false;

	/*
	 * take a ref on the afu so that it stays alive at-least till
	 * this context is reclaimed inside reclaim_ctx.
	 */
	cxl_afu_get(afu);
	return 0;
}

static int cxl_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
	struct cxl_context *ctx = vma->vm_file->private_data;
	unsigned long address = (unsigned long)vmf->virtual_address;
	u64 area, offset;

	offset = vmf->pgoff << PAGE_SHIFT;

	pr_devel("%s: pe: %i address: 0x%lx offset: 0x%llx\n",
			__func__, ctx->pe, address, offset);

	if (ctx->afu->current_mode == CXL_MODE_DEDICATED) {
		area = ctx->afu->psn_phys;
		if (offset >= ctx->afu->adapter->ps_size)
			return VM_FAULT_SIGBUS;
	} else {
		area = ctx->psn_phys;
		if (offset >= ctx->psn_size)
			return VM_FAULT_SIGBUS;
	}

	mutex_lock(&ctx->status_mutex);

	if (ctx->status != STARTED) {
		mutex_unlock(&ctx->status_mutex);
		pr_devel("%s: Context not started, failing problem state access\n", __func__);
		if (ctx->mmio_err_ff) {
			if (!ctx->ff_page) {
				ctx->ff_page = alloc_page(GFP_USER);
				if (!ctx->ff_page)
					return VM_FAULT_OOM;
				memset(page_address(ctx->ff_page), 0xff, PAGE_SIZE);
			}
			get_page(ctx->ff_page);
			vmf->page = ctx->ff_page;
			vma->vm_page_prot = pgprot_cached(vma->vm_page_prot);
			return 0;
		}
		return VM_FAULT_SIGBUS;
	}

	vm_insert_pfn(vma, address, (area + offset) >> PAGE_SHIFT);

	mutex_unlock(&ctx->status_mutex);

	return VM_FAULT_NOPAGE;
}

static const struct vm_operations_struct cxl_mmap_vmops = {
	.fault = cxl_mmap_fault,
};

/*
 * Map a per-context mmio space into the given vma.
 */
int cxl_context_iomap(struct cxl_context *ctx, struct vm_area_struct *vma)
{
	u64 start = vma->vm_pgoff << PAGE_SHIFT;
	u64 len = vma->vm_end - vma->vm_start;

	if (ctx->afu->current_mode == CXL_MODE_DEDICATED) {
		if (start + len > ctx->afu->adapter->ps_size)
			return -EINVAL;
	} else {
		if (start + len > ctx->psn_size)
			return -EINVAL;
	}

	if (ctx->afu->current_mode != CXL_MODE_DEDICATED) {
		/* make sure there is a valid per process space for this AFU */
		if ((ctx->master && !ctx->afu->psa) || (!ctx->afu->pp_psa)) {
			pr_devel("AFU doesn't support mmio space\n");
			return -EINVAL;
		}

		/* Can't mmap until the AFU is enabled */
		if (!ctx->afu->enabled)
			return -EBUSY;
	}

	pr_devel("%s: mmio physical: %llx pe: %i master:%i\n", __func__,
		 ctx->psn_phys, ctx->pe , ctx->master);

	vma->vm_flags |= VM_IO | VM_PFNMAP;
	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
	vma->vm_ops = &cxl_mmap_vmops;
	return 0;
}

/*
 * Detach a context from the hardware. This disables interrupts and doesn't
 * return until all outstanding interrupts for this context have completed. The
 * hardware should no longer access *ctx after this has returned.
 */
int __detach_context(struct cxl_context *ctx)
{
	enum cxl_context_status status;

	mutex_lock(&ctx->status_mutex);
	status = ctx->status;
	ctx->status = CLOSED;
	mutex_unlock(&ctx->status_mutex);
	if (status != STARTED)
		return -EBUSY;

	/* Only warn if we detached while the link was OK.
	 * If detach fails when hw is down, we don't care.
	 */
	WARN_ON(cxl_ops->detach_process(ctx) &&
		cxl_ops->link_ok(ctx->afu->adapter, ctx->afu));
	flush_work(&ctx->fault_work); /* Only needed for dedicated process */

	/*
	 * Wait until no further interrupts are presented by the PSL
	 * for this context.
	 */
	if (cxl_ops->irq_wait)
		cxl_ops->irq_wait(ctx);

	/* release the reference to the group leader and mm handling pid */
	put_pid(ctx->pid);
	put_pid(ctx->glpid);

	cxl_ctx_put();
	return 0;
}

/*
 * Detach the given context from the AFU. This doesn't actually
 * free the context but it should stop the context running in hardware
 * (ie. prevent this context from generating any further interrupts
 * so that it can be freed).
 */
void cxl_context_detach(struct cxl_context *ctx)
{
	int rc;

	rc = __detach_context(ctx);
	if (rc)
		return;

	afu_release_irqs(ctx, ctx);
	wake_up_all(&ctx->wq);
}

/*
 * Detach all contexts on the given AFU.
 */
void cxl_context_detach_all(struct cxl_afu *afu)
{
	struct cxl_context *ctx;
	int tmp;

	mutex_lock(&afu->contexts_lock);
	idr_for_each_entry(&afu->contexts_idr, ctx, tmp) {
		/*
		 * Anything done in here needs to be setup before the IDR is
		 * created and torn down after the IDR removed
		 */
		cxl_context_detach(ctx);

		/*
		 * We are force detaching - remove any active PSA mappings so
		 * userspace cannot interfere with the card if it comes back.
		 * Easiest way to exercise this is to unbind and rebind the
		 * driver via sysfs while it is in use.
		 */
		mutex_lock(&ctx->mapping_lock);
		if (ctx->mapping)
			unmap_mapping_range(ctx->mapping, 0, 0, 1);
		mutex_unlock(&ctx->mapping_lock);
	}
	mutex_unlock(&afu->contexts_lock);
}

static void reclaim_ctx(struct rcu_head *rcu)
{
	struct cxl_context *ctx = container_of(rcu, struct cxl_context, rcu);

	free_page((u64)ctx->sstp);
	if (ctx->ff_page)
		__free_page(ctx->ff_page);
	ctx->sstp = NULL;
	if (ctx->kernelapi)
		kfree(ctx->mapping);

	kfree(ctx->irq_bitmap);

	/* Drop ref to the afu device taken during cxl_context_init */
	cxl_afu_put(ctx->afu);

	kfree(ctx);
}

void cxl_context_free(struct cxl_context *ctx)
{
	mutex_lock(&ctx->afu->contexts_lock);
	idr_remove(&ctx->afu->contexts_idr, ctx->pe);
	mutex_unlock(&ctx->afu->contexts_lock);
	call_rcu(&ctx->rcu, reclaim_ctx);
}
Commit	Line	Data
f204e0b8 IM	1	/*
	2	* Copyright 2014 IBM Corp.
	3	*
	4	* This program is free software; you can redistribute it and/or
	5	* modify it under the terms of the GNU General Public License
	6	* as published by the Free Software Foundation; either version
	7	* 2 of the License, or (at your option) any later version.
	8	*/
	9
	10	#include <linux/module.h>
	11	#include <linux/kernel.h>
	12	#include <linux/bitmap.h>
	13	#include <linux/sched.h>
	14	#include <linux/pid.h>
	15	#include <linux/fs.h>
	16	#include <linux/mm.h>
	17	#include <linux/debugfs.h>
	18	#include <linux/slab.h>
	19	#include <linux/idr.h>
	20	#include <asm/cputable.h>
	21	#include <asm/current.h>
	22	#include <asm/copro.h>
	23
	24	#include "cxl.h"
	25
	26	/*
	27	* Allocates space for a CXL context.
	28	*/
	29	struct cxl_context *cxl_context_alloc(void)
	30	{
	31	return kzalloc(sizeof(struct cxl_context), GFP_KERNEL);
	32	}
	33
	34	/*
	35	* Initialises a CXL context.
	36	*/
b123429e IM	37	int cxl_context_init(struct cxl_context ctx, struct cxl_afu afu, bool master,
b123429e IM	38	struct address_space *mapping)
f204e0b8 IM	39	{
	40	int i;
	41
	42	spin_lock_init(&ctx->sste_lock);
	43	ctx->afu = afu;
	44	ctx->master = master;
7b8ad495	45	ctx->pid = ctx->glpid = NULL; /* Set in start work ioctl */
b123429e IM	46	mutex_init(&ctx->mapping_lock);
b123429e IM	47	ctx->mapping = mapping;
f204e0b8 IM	48
	49	/*
	50	* Allocate the segment table before we put it in the IDR so that we
	51	* can always access it when dereferenced from IDR. For the same
	52	* reason, the segment table is only destroyed after the context is
	53	* removed from the IDR. Access to this in the IOCTL is protected by
	54	* Linux filesytem symantics (can't IOCTL until open is complete).
	55	*/
	56	i = cxl_alloc_sst(ctx);
	57	if (i)
	58	return i;
	59
	60	INIT_WORK(&ctx->fault_work, cxl_handle_fault);
	61
	62	init_waitqueue_head(&ctx->wq);
	63	spin_lock_init(&ctx->lock);
	64
	65	ctx->irq_bitmap = NULL;
	66	ctx->pending_irq = false;
	67	ctx->pending_fault = false;
	68	ctx->pending_afu_err = false;
	69
f5c9df9a	70	INIT_LIST_HEAD(&ctx->irq_names);
cbce0917	71	INIT_LIST_HEAD(&ctx->extra_irq_contexts);
f5c9df9a	72
f204e0b8 IM	73	/*
	74	* When we have to destroy all contexts in cxl_context_detach_all() we
	75	* end up with afu_release_irqs() called from inside a
	76	* idr_for_each_entry(). Hence we need to make sure that anything
	77	* dereferenced from this IDR is ok before we allocate the IDR here.
	78	* This clears out the IRQ ranges to ensure this.
	79	*/
	80	for (i = 0; i < CXL_IRQ_RANGES; i++)
	81	ctx->irqs.range[i] = 0;
	82
	83	mutex_init(&ctx->status_mutex);
	84
	85	ctx->status = OPENED;
	86
	87	/*
	88	* Allocating IDR! We better make sure everything's setup that
	89	* dereferences from it.
	90	*/
ee41d11d	91	mutex_lock(&afu->contexts_lock);
f204e0b8	92	idr_preload(GFP_KERNEL);
f67a6722 IM	93	i = idr_alloc(&ctx->afu->contexts_idr, ctx,
f67a6722 IM	94	ctx->afu->adapter->native->sl_ops->min_pe,
f204e0b8	95	ctx->afu->num_procs, GFP_NOWAIT);
f204e0b8	96	idr_preload_end();
ee41d11d	97	mutex_unlock(&afu->contexts_lock);
f204e0b8 IM	98	if (i < 0)
	99	return i;
	100
	101	ctx->pe = i;
14baf4d9	102	if (cpu_has_feature(CPU_FTR_HVMODE)) {
cbffa3a5	103	ctx->elem = &ctx->afu->native->spa[i];
14baf4d9 CL	104	ctx->external_pe = ctx->pe;
	105	} else {
	106	ctx->external_pe = -1; /* assigned when attaching */
	107	}
f204e0b8	108	ctx->pe_inserted = false;
1b5df59e VJ	109
	110	/*
	111	* take a ref on the afu so that it stays alive at-least till
	112	* this context is reclaimed inside reclaim_ctx.
	113	*/
	114	cxl_afu_get(afu);
f204e0b8 IM	115	return 0;
	116	}
	117
0712dc7e IM	118	static int cxl_mmap_fault(struct vm_area_struct vma, struct vm_fault vmf)
	119	{
	120	struct cxl_context *ctx = vma->vm_file->private_data;
	121	unsigned long address = (unsigned long)vmf->virtual_address;
	122	u64 area, offset;
	123
	124	offset = vmf->pgoff << PAGE_SHIFT;
	125
	126	pr_devel("%s: pe: %i address: 0x%lx offset: 0x%llx\n",
	127	__func__, ctx->pe, address, offset);
	128
	129	if (ctx->afu->current_mode == CXL_MODE_DEDICATED) {
	130	area = ctx->afu->psn_phys;
10a5894f	131	if (offset >= ctx->afu->adapter->ps_size)
0712dc7e IM	132	return VM_FAULT_SIGBUS;
	133	} else {
	134	area = ctx->psn_phys;
10a5894f	135	if (offset >= ctx->psn_size)
0712dc7e IM	136	return VM_FAULT_SIGBUS;
	137	}
	138
	139	mutex_lock(&ctx->status_mutex);
	140
	141	if (ctx->status != STARTED) {
	142	mutex_unlock(&ctx->status_mutex);
	143	pr_devel("%s: Context not started, failing problem state access\n", __func__);
d9232a3d IM	144	if (ctx->mmio_err_ff) {
	145	if (!ctx->ff_page) {
	146	ctx->ff_page = alloc_page(GFP_USER);
	147	if (!ctx->ff_page)
	148	return VM_FAULT_OOM;
	149	memset(page_address(ctx->ff_page), 0xff, PAGE_SIZE);
	150	}
	151	get_page(ctx->ff_page);
	152	vmf->page = ctx->ff_page;
	153	vma->vm_page_prot = pgprot_cached(vma->vm_page_prot);
	154	return 0;
	155	}
0712dc7e IM	156	return VM_FAULT_SIGBUS;
	157	}
	158
	159	vm_insert_pfn(vma, address, (area + offset) >> PAGE_SHIFT);
	160
	161	mutex_unlock(&ctx->status_mutex);
	162
	163	return VM_FAULT_NOPAGE;
	164	}
	165
	166	static const struct vm_operations_struct cxl_mmap_vmops = {
	167	.fault = cxl_mmap_fault,
	168	};
	169
f204e0b8 IM	170	/*
	171	* Map a per-context mmio space into the given vma.
	172	*/
	173	int cxl_context_iomap(struct cxl_context ctx, struct vm_area_struct vma)
	174	{
5caaf534	175	u64 start = vma->vm_pgoff << PAGE_SHIFT;
f204e0b8	176	u64 len = vma->vm_end - vma->vm_start;
5caaf534 IM	177
	178	if (ctx->afu->current_mode == CXL_MODE_DEDICATED) {
	179	if (start + len > ctx->afu->adapter->ps_size)
	180	return -EINVAL;
	181	} else {
	182	if (start + len > ctx->psn_size)
	183	return -EINVAL;
	184	}
f204e0b8	185
0712dc7e IM	186	if (ctx->afu->current_mode != CXL_MODE_DEDICATED) {
	187	/* make sure there is a valid per process space for this AFU */
	188	if ((ctx->master && !ctx->afu->psa) \|\| (!ctx->afu->pp_psa)) {
	189	pr_devel("AFU doesn't support mmio space\n");
	190	return -EINVAL;
	191	}
f204e0b8	192
0712dc7e IM	193	/* Can't mmap until the AFU is enabled */
	194	if (!ctx->afu->enabled)
	195	return -EBUSY;
f204e0b8 IM	196	}
f204e0b8 IM	197
f204e0b8 IM	198	pr_devel("%s: mmio physical: %llx pe: %i master:%i\n", __func__,
	199	ctx->psn_phys, ctx->pe , ctx->master);
	200
0712dc7e	201	vma->vm_flags \|= VM_IO \| VM_PFNMAP;
f204e0b8	202	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
0712dc7e IM	203	vma->vm_ops = &cxl_mmap_vmops;
0712dc7e IM	204	return 0;
f204e0b8 IM	205	}
	206
	207	/*
	208	* Detach a context from the hardware. This disables interrupts and doesn't
	209	* return until all outstanding interrupts for this context have completed. The
	210	* hardware should no longer access *ctx after this has returned.
	211	*/
eda3693c	212	int __detach_context(struct cxl_context *ctx)
f204e0b8 IM	213	{
	214	enum cxl_context_status status;
	215
	216	mutex_lock(&ctx->status_mutex);
	217	status = ctx->status;
	218	ctx->status = CLOSED;
	219	mutex_unlock(&ctx->status_mutex);
	220	if (status != STARTED)
eda3693c	221	return -EBUSY;
f204e0b8	222
0b3f9c75 DA	223	/* Only warn if we detached while the link was OK.
	224	* If detach fails when hw is down, we don't care.
	225	*/
5be587b1	226	WARN_ON(cxl_ops->detach_process(ctx) &&
0d400f77	227	cxl_ops->link_ok(ctx->afu->adapter, ctx->afu));
7bb5d91a	228	flush_work(&ctx->fault_work); /* Only needed for dedicated process */
7b8ad495	229
2bc79ffc MN	230	/*
	231	* Wait until no further interrupts are presented by the PSL
	232	* for this context.
	233	*/
	234	if (cxl_ops->irq_wait)
	235	cxl_ops->irq_wait(ctx);
	236
7b8ad495	237	/* release the reference to the group leader and mm handling pid */
7bb5d91a	238	put_pid(ctx->pid);
7b8ad495 VJ	239	put_pid(ctx->glpid);
7b8ad495 VJ	240
7bb5d91a	241	cxl_ctx_put();
eda3693c	242	return 0;
f204e0b8 IM	243	}
	244
	245	/*
	246	* Detach the given context from the AFU. This doesn't actually
	247	* free the context but it should stop the context running in hardware
	248	* (ie. prevent this context from generating any further interrupts
	249	* so that it can be freed).
	250	*/
	251	void cxl_context_detach(struct cxl_context *ctx)
	252	{
eda3693c MN	253	int rc;
	254
	255	rc = __detach_context(ctx);
	256	if (rc)
	257	return;
	258
	259	afu_release_irqs(ctx, ctx);
eda3693c	260	wake_up_all(&ctx->wq);
f204e0b8 IM	261	}
	262
	263	/*
	264	* Detach all contexts on the given AFU.
	265	*/
	266	void cxl_context_detach_all(struct cxl_afu *afu)
	267	{
	268	struct cxl_context *ctx;
	269	int tmp;
	270
ee41d11d IM	271	mutex_lock(&afu->contexts_lock);
ee41d11d IM	272	idr_for_each_entry(&afu->contexts_idr, ctx, tmp) {
f204e0b8 IM	273	/*
	274	* Anything done in here needs to be setup before the IDR is
	275	* created and torn down after the IDR removed
	276	*/
eda3693c	277	cxl_context_detach(ctx);
0712dc7e IM	278
	279	/*
	280	* We are force detaching - remove any active PSA mappings so
	281	* userspace cannot interfere with the card if it comes back.
	282	* Easiest way to exercise this is to unbind and rebind the
	283	* driver via sysfs while it is in use.
	284	*/
	285	mutex_lock(&ctx->mapping_lock);
	286	if (ctx->mapping)
	287	unmap_mapping_range(ctx->mapping, 0, 0, 1);
	288	mutex_unlock(&ctx->mapping_lock);
ee41d11d IM	289	}
ee41d11d IM	290	mutex_unlock(&afu->contexts_lock);
f204e0b8 IM	291	}
f204e0b8 IM	292
8ac75b96	293	static void reclaim_ctx(struct rcu_head *rcu)
f204e0b8	294	{
8ac75b96	295	struct cxl_context *ctx = container_of(rcu, struct cxl_context, rcu);
f204e0b8 IM	296
f204e0b8 IM	297	free_page((u64)ctx->sstp);
d9232a3d IM	298	if (ctx->ff_page)
d9232a3d IM	299	__free_page(ctx->ff_page);
f204e0b8	300	ctx->sstp = NULL;
55e07668 IM	301	if (ctx->kernelapi)
55e07668 IM	302	kfree(ctx->mapping);
f204e0b8	303
1050e689	304	kfree(ctx->irq_bitmap);
52adee58	305
1b5df59e VJ	306	/* Drop ref to the afu device taken during cxl_context_init */
	307	cxl_afu_put(ctx->afu);
	308
f204e0b8 IM	309	kfree(ctx);
f204e0b8 IM	310	}
8ac75b96 IM	311
	312	void cxl_context_free(struct cxl_context *ctx)
	313	{
	314	mutex_lock(&ctx->afu->contexts_lock);
	315	idr_remove(&ctx->afu->contexts_idr, ctx->pe);
	316	mutex_unlock(&ctx->afu->contexts_lock);
	317	call_rcu(&ctx->rcu, reclaim_ctx);
	318	}