Skip to contentSkip to conversation

iommu/vt-d: Use ida for domain ID management

PATCH Superseded Lu Baolu 4 participants 14 messages Apr 23, 2025 – Apr 27, 2025

Subsystems (3)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Lu Baolu

author maintainer
#0 lore raw

cover iommu/vt-d: Use ida for domain ID management

This converts the Intel iommu driver's domain ID management from a fixed-size bitmap to the dynamic ida allocator. This improves memory efficiency by only allocating resources for the domain IDs actually in use, rather than the maximum possible number.

The also includes necessary cleanups after the ida conversion, including locking adjustment for the ida and code simplification.

Please review and comment.

Lu Baolu (3):
iommu/vt-d: Use ida to manage domain id
iommu/vt-d: Replace spin_lock with mutex to protect domain ida iommu/vt-d: Simplify domain_attach_iommu()

 drivers/iommu/intel/dmar.c  |   3 +
 drivers/iommu/intel/iommu.c | 113 ++++++++----------------------------
 drivers/iommu/intel/iommu.h |  20 +++++--
 3 files changed, 44 insertions(+), 92 deletions(-)
Signature · 1 line
2.43.0
author maintainer
#1 lore raw

PATCH 1/3 iommu/vt-d: Use ida to manage domain id

Switch the intel iommu driver to use the ida mechanism for managing domain IDs, replacing the previous fixed-size bitmap.

The previous approach allocated a bitmap large enough to cover the maximum number of domain IDs supported by the hardware, regardless of the actual number of domains in use. This led to unnecessary memory consumption, especially on systems supporting a large number of iommu units but only utilizing a small number of domain IDs.

The ida allocator dynamically manages the allocation and freeing of integer IDs, only consuming memory for the IDs that are currently in use. This significantly optimizes memory usage compared to the fixed-size bitmap.

---
 drivers/iommu/intel/dmar.c  |  2 +
 drivers/iommu/intel/iommu.c | 80 ++++++++-----------------------------
 drivers/iommu/intel/iommu.h | 18 +++++++--
 3 files changed, 33 insertions(+), 67 deletions(-)
drivers/iommu/intel/dmar.c +2 -0
index e540092d664d..0afba7434a5c 100644
@@ -1099,6 +1099,8 @@ static int alloc_iommu(struct dmar_drhd_unit *drhd)
 	spin_lock_init(&iommu->device_rbtree_lock);
 	mutex_init(&iommu->iopf_lock);
 	iommu->node = NUMA_NO_NODE;
+	spin_lock_init(&iommu->lock);
+	ida_init(&iommu->domain_ida);
 
 	ver = readl(iommu->reg + DMAR_VER_REG);
 	pr_info("%s: reg_base_addr %llx ver %d:%d cap %llx ecap %llx\n",
drivers/iommu/intel/iommu.c +17 -63

150 lines not on this page · raw message

drivers/iommu/intel/iommu.h +14 -4
index c4916886da5a..993232292400 100644
@@ -722,7 +722,7 @@ struct intel_iommu {
 	unsigned char	name[16];    /* Device Name */
 
 #ifdef CONFIG_INTEL_IOMMU
-	unsigned long 	*domain_ids; /* bitmap of domains */
+	struct ida	domain_ida; /* domain id allocator */
 	unsigned long	*copied_tables; /* bitmap of copied tables */
 	spinlock_t	lock; /* protect context, domain ids */
 	struct root_entry *root_entry; /* virtual address */
@@ -809,11 +809,21 @@ static inline struct dmar_domain *to_dmar_domain(struct iommu_domain *dom)
 }
 
 /*
- * Domain ID reserved for pasid entries programmed for first-level
- * only and pass-through transfer modes.
+ * Domain ID 0 and 1 are reserved:
+ *
+ * If Caching mode is set, then invalid translations are tagged
+ * with domain-id 0, hence we need to pre-allocate it. We also
+ * use domain-id 0 as a marker for non-allocated domain-id, so
+ * make sure it is not used for a real domain.
+ *
+ * Vt-d spec rev3.0 (section 6.2.3.1) requires that each pasid
+ * entry for first-level or pass-through translation modes should
+ * be programmed with a domain id different from those used for
+ * second-level or nested translation. We reserve a domain id for
+ * this purpose. This domain id is also used for identity domain
+ * in legacy mode.
  */
 #define FLPT_DEFAULT_DID		1
-#define NUM_RESERVED_DID		2
 
 /* Retrieve the domain ID which has allocated to the domain */
 static inline u16
Signature · 1 line
2.43.0
Reviewed-by on v1 1/3
#2 lore raw

RE: [PATCH 1/3] iommu/vt-d: Use ida to manage domain id

drivers/iommu/intel/iommu.c static void disable_dmar_iommu(struct intel_iommu *iommu) · line 1307

Quoted 10 lines

From: Lu Baolu <[email protected]>
Sent: Wednesday, April 23, 2025 11:10 AM

static void free_dmar_iommu(struct intel_iommu *iommu) {

drivers/iommu/intel/iommu.c +1 -4
-	if (iommu->domain_ids) {
-		bitmap_free(iommu->domain_ids);
-		iommu->domain_ids = NULL;
-	}
+	ida_destroy(&iommu->domain_ida);

since ida_init() is in alloc_iommu() now, the destroy can be moved to free_iommu().

drivers/iommu/intel/iommu.c int domain_attach_iommu(struct dmar_domain · line 1362

Quoted 11 lines
drivers/iommu/intel/iommu.c +2 -3
@@ -1399,14 +1356,13 @@ int domain_attach_iommu(struct dmar_domain
 *domain, struct intel_iommu *iommu)
 		return 0;
 	}
 
-	ndomains = cap_ndoms(iommu->cap);
-	num = find_first_zero_bit(iommu->domain_ids, ndomains);
-	if (num >= ndomains) {
+	num = ida_alloc_range(&iommu->domain_ida, FLPT_DEFAULT_DID +
 1,
+			      cap_ndoms(iommu->cap) - 1, GFP_ATOMIC);

let's define a macro e.g. FIRST_DID for min.

author maintainer
#3 lore raw

Re: [PATCH 1/3] iommu/vt-d: Use ida to manage domain id

Quoted 13 lines · Kevin

Review

Quoted 10 lines

From: Lu Baolu <[email protected]>
Sent: Wednesday, April 23, 2025 11:10 AM

  static void free_dmar_iommu(struct intel_iommu *iommu)
  {
(quoted diff) +1 -4
-	if (iommu->domain_ids) {
-		bitmap_free(iommu->domain_ids);
-		iommu->domain_ids = NULL;
-	}
+	ida_destroy(&iommu->domain_ida);

since ida_init() is in alloc_iommu() now, the destroy can be moved to free_iommu().

Sure.

drivers/iommu/intel/dmar.c +1 -0
index e25e8d3dedcb..9e17e8e56308 100644
@@ -1198,6 +1198,7 @@ static void free_iommu(struct intel_iommu *iommu)
        if (iommu->reg)
                unmap_iommu(iommu);
 
+       ida_destroy(&iommu->domain_ida);
        ida_free(&dmar_seq_ids, iommu->seq_id);
        kfree(iommu);
 }
drivers/iommu/intel/iommu.c +0 -2
index f784df122d7a..0002511c561f 100644
@@ -1304,8 +1304,6 @@ static void disable_dmar_iommu(struct intel_iommu *iommu)
 
 static void free_dmar_iommu(struct intel_iommu *iommu)
 {
-       ida_destroy(&iommu->domain_ida);
-
        if (iommu->copied_tables) {
                bitmap_free(iommu->copied_tables);
                iommu->copied_tables = NULL;
Quoted 6 lines
Quoted 6 lines
(quoted diff) +0 -0
@@ -1399,14 +1356,13 @@ int domain_attach_iommu(struct dmar_domain
 *domain, struct intel_iommu *iommu)
  		return 0;
  	}
 
Quoted 8 lines

Review int domain_attach_iommu(struct dmar_domain · line 1362

Quoted 6 lines
(quoted diff) +2 -3
-	ndomains = cap_ndoms(iommu->cap);
-	num = find_first_zero_bit(iommu->domain_ids, ndomains);
-	if (num >= ndomains) {
+	num = ida_alloc_range(&iommu->domain_ida, FLPT_DEFAULT_DID +
 1,
+			      cap_ndoms(iommu->cap) - 1, GFP_ATOMIC);

let's define a macro e.g. FIRST_DID for min.

Okay, how about IDA_START_DID?

Thanks,
baolu

Quoted 2 lines
author maintainer
#4 lore raw

Re: [PATCH 1/3] iommu/vt-d: Use ida to manage domain id

In reply to Lu Baolu

drivers/iommu/intel/iommu.c static int copy_context_table(struct intel_iommu *iommu, · line 2001

Quoted 6 lines · Lu Baolu
drivers/iommu/intel/iommu.c +1 -1
@@ -2042,7 +1998,7 @@ static int copy_context_table(struct intel_iommu *iommu,
 
  		did = context_domain_id(&ce);
  		if (did >= 0 && did < cap_ndoms(iommu->cap))
-			set_bit(did, iommu->domain_ids);
+			ida_alloc_range(&iommu->domain_ida, did, did + 1, GFP_KERNEL);

ida allocation range is inclusive, thus here it should be,

ida_alloc_range(&iommu->domain_ida, did, did, GFP_KERNEL);

drivers/iommu/intel/iommu.c static int copy_context_table(struct intel_iommu *iommu, · line 2004

Quoted 3 lines
drivers/iommu/intel/iommu.c +0 -0
  		set_context_copied(iommu, bus, devfn);
  		new_ce[idx] = ce;

Thanks,
baolu

author maintainer
#5 lore raw

PATCH 2/3 iommu/vt-d: Replace spin_lock with mutex to protect domain ida

In reply to Lu Baolu

The domain ID allocator is currently protected by a spin_lock. However, ida_alloc_range can potentially block if it needs to allocate memory to grow its internal structures.

Replace the spin_lock with a mutex which allows sleep on block. Thus, the memory allocation flags can be updated from GFP_ATOMIC to GFP_KERNEL to allow blocking memory allocations if necessary.

Introduce a new mutex, did_lock, specifically for protecting the domain ida. The existing spinlock will remain for protecting other intel_iommu fields.

---
 drivers/iommu/intel/dmar.c  |  1 +
 drivers/iommu/intel/iommu.c | 12 ++++--------
 drivers/iommu/intel/iommu.h |  2 ++
 3 files changed, 7 insertions(+), 8 deletions(-)
drivers/iommu/intel/dmar.c +1 -0
index 0afba7434a5c..e25e8d3dedcb 100644
@@ -1101,6 +1101,7 @@ static int alloc_iommu(struct dmar_drhd_unit *drhd)
 	iommu->node = NUMA_NO_NODE;
 	spin_lock_init(&iommu->lock);
 	ida_init(&iommu->domain_ida);
+	mutex_init(&iommu->did_lock);
 
 	ver = readl(iommu->reg + DMAR_VER_REG);
 	pr_info("%s: reg_base_addr %llx ver %d:%d cap %llx ecap %llx\n",
drivers/iommu/intel/iommu.c +4 -8
index 76e170922149..3a9ea0ad2cd3 100644
@@ -1347,17 +1347,16 @@ int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
 	if (!info)
 		return -ENOMEM;
 
-	spin_lock(&iommu->lock);
+	guard(mutex)(&iommu->did_lock);
 	curr = xa_load(&domain->iommu_array, iommu->seq_id);
 	if (curr) {
 		curr->refcnt++;
-		spin_unlock(&iommu->lock);
 		kfree(info);
 		return 0;
 	}
 
 	num = ida_alloc_range(&iommu->domain_ida, FLPT_DEFAULT_DID + 1,
-			      cap_ndoms(iommu->cap) - 1, GFP_ATOMIC);
+			      cap_ndoms(iommu->cap) - 1, GFP_KERNEL);
 	if (num < 0) {
 		pr_err("%s: No free domain ids\n", iommu->name);
 		goto err_unlock;
@@ -1367,19 +1366,17 @@ int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
 	info->did	= num;
 	info->iommu	= iommu;
 	curr = xa_cmpxchg(&domain->iommu_array, iommu->seq_id,
-			  NULL, info, GFP_ATOMIC);
+			  NULL, info, GFP_KERNEL);
 	if (curr) {
 		ret = xa_err(curr) ? : -EBUSY;
 		goto err_clear;
 	}
 
-	spin_unlock(&iommu->lock);
 	return 0;
 
 err_clear:
 	ida_free(&iommu->domain_ida, info->did);
 err_unlock:
-	spin_unlock(&iommu->lock);
 	kfree(info);
 	return ret;
 }
@@ -1391,7 +1388,7 @@ void domain_detach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
 	if (domain->domain.type == IOMMU_DOMAIN_SVA)
 		return;
 
-	spin_lock(&iommu->lock);
+	guard(mutex)(&iommu->did_lock);
 	info = xa_load(&domain->iommu_array, iommu->seq_id);
 	if (--info->refcnt == 0) {
 		ida_free(&iommu->domain_ida, info->did);
@@ -1399,7 +1396,6 @@ void domain_detach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
 		domain->nid = NUMA_NO_NODE;
 		kfree(info);
 	}
-	spin_unlock(&iommu->lock);
 }
 
 static void domain_exit(struct dmar_domain *domain)
drivers/iommu/intel/iommu.h +2 -0
index 993232292400..1e98c10ca553 100644
@@ -722,6 +722,8 @@ struct intel_iommu {
 	unsigned char	name[16];    /* Device Name */
 
 #ifdef CONFIG_INTEL_IOMMU
+	/* mutex to protect domain_ida */
+	struct mutex	did_lock;
 	struct ida	domain_ida; /* domain id allocator */
 	unsigned long	*copied_tables; /* bitmap of copied tables */
 	spinlock_t	lock; /* protect context, domain ids */
Signature · 1 line
2.43.0
Reviewed-by on v1 2/3
#6 lore raw

RE: [PATCH 2/3] iommu/vt-d: Replace spin_lock with mutex to protect domain ida

Quoted 12 lines

From: Lu Baolu <[email protected]>
Sent: Wednesday, April 23, 2025 11:10 AM

The domain ID allocator is currently protected by a spin_lock. However, ida_alloc_range can potentially block if it needs to allocate memory to grow its internal structures.

Replace the spin_lock with a mutex which allows sleep on block. Thus, the memory allocation flags can be updated from GFP_ATOMIC to GFP_KERNEL
to allow blocking memory allocations if necessary.

Quoted 5 lines

Introduce a new mutex, did_lock, specifically for protecting the domain ida. The existing spinlock will remain for protecting other intel_iommu fields.

author maintainer
#7 lore raw

PATCH 3/3 iommu/vt-d: Simplify domain_attach_iommu()

In reply to Lu Baolu

Use the __free(kfree) attribute with kzalloc() to automatically handle the freeing of the allocated struct iommu_domain_info on error or early exit paths, eliminating the need for explicit kfree() calls in error handling branches.

---
 drivers/iommu/intel/iommu.c | 29 ++++++++---------------------
 1 file changed, 8 insertions(+), 21 deletions(-)
drivers/iommu/intel/iommu.c +8 -21
index 3a9ea0ad2cd3..12382c85495f 100644
@@ -1337,13 +1337,14 @@ static bool first_level_by_default(struct intel_iommu *iommu)
 
 int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
 {
-	struct iommu_domain_info *info, *curr;
-	int num, ret = -ENOSPC;
+	struct iommu_domain_info *curr;
+	int num;
 
 	if (domain->domain.type == IOMMU_DOMAIN_SVA)
 		return 0;
 
-	info = kzalloc(sizeof(*info), GFP_KERNEL);
+	struct iommu_domain_info *info __free(kfree) =
+		kzalloc(sizeof(*info), GFP_KERNEL);
 	if (!info)
 		return -ENOMEM;
 
@@ -1351,34 +1352,20 @@ int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
 	curr = xa_load(&domain->iommu_array, iommu->seq_id);
 	if (curr) {
 		curr->refcnt++;
-		kfree(info);
 		return 0;
 	}
 
 	num = ida_alloc_range(&iommu->domain_ida, FLPT_DEFAULT_DID + 1,
 			      cap_ndoms(iommu->cap) - 1, GFP_KERNEL);
-	if (num < 0) {
-		pr_err("%s: No free domain ids\n", iommu->name);
-		goto err_unlock;
-	}
+	if (num < 0)
+		return num;
 
 	info->refcnt	= 1;
 	info->did	= num;
 	info->iommu	= iommu;
-	curr = xa_cmpxchg(&domain->iommu_array, iommu->seq_id,
-			  NULL, info, GFP_KERNEL);
-	if (curr) {
-		ret = xa_err(curr) ? : -EBUSY;
-		goto err_clear;
-	}
 
-	return 0;
-
-err_clear:
-	ida_free(&iommu->domain_ida, info->did);
-err_unlock:
-	kfree(info);
-	return ret;
+	return xa_err(xa_store(&domain->iommu_array, iommu->seq_id,
+			       no_free_ptr(info), GFP_KERNEL));
 }
 
 void domain_detach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
Signature · 1 line
2.43.0
#8 lore raw

RE: [PATCH 3/3] iommu/vt-d: Simplify domain_attach_iommu()

drivers/iommu/intel/iommu.c int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu) · line 1361

Quoted 8 lines

From: Lu Baolu <[email protected]>
Sent: Wednesday, April 23, 2025 11:10 AM

num = ida_alloc_range(&iommu->domain_ida, FLPT_DEFAULT_DID +
1,

drivers/iommu/intel/iommu.c +0 -2
 			      cap_ndoms(iommu->cap) - 1, GFP_KERNEL);
-	if (num < 0) {
-		pr_err("%s: No free domain ids\n", iommu->name);

this error message could be kept.

Quoted 15 lines
drivers/iommu/intel/iommu.c +2 -8
-		goto err_unlock;
-	}
+	if (num < 0)
+		return num;
 
 	info->refcnt	= 1;
 	info->did	= num;
 	info->iommu	= iommu;
-	curr = xa_cmpxchg(&domain->iommu_array, iommu->seq_id,
-			  NULL, info, GFP_KERNEL);
-	if (curr) {
-		ret = xa_err(curr) ? : -EBUSY;
-		goto err_clear;
-	}
 

drivers/iommu/intel/iommu.c int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu) · line 1369

Quoted 10 lines
drivers/iommu/intel/iommu.c +2 -7
-	return 0;
-
-err_clear:
-	ida_free(&iommu->domain_ida, info->did);
-err_unlock:
-	kfree(info);
-	return ret;
+	return xa_err(xa_store(&domain->iommu_array, iommu->seq_id,
+			       no_free_ptr(info), GFP_KERNEL));
 }

no_free_ptr() should be used before successful return. Here xa_store() could return error but at that point no auto free as no_free_ptr() already changes 'info' to NULL. then memory leak.

author maintainer
#9 lore raw

Re: [PATCH 3/3] iommu/vt-d: Simplify domain_attach_iommu()

Quoted 10 lines · Kevin

drivers/iommu/intel/iommu.c int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu) · line 1361

Quoted 8 lines

From: Lu Baolu <[email protected]>
Sent: Wednesday, April 23, 2025 11:10 AM

num = ida_alloc_range(&iommu->domain_ida, FLPT_DEFAULT_DID +
1,
cap_ndoms(iommu->cap) - 1, GFP_KERNEL);

drivers/iommu/intel/iommu.c +0 -2
-	if (num < 0) {
-		pr_err("%s: No free domain ids\n", iommu->name);

this error message could be kept.

Okay.

Quoted 16 lines
Quoted 16 lines
drivers/iommu/intel/iommu.c +2 -8
-		goto err_unlock;
-	}
+	if (num < 0)
+		return num;
 
  	info->refcnt	= 1;
  	info->did	= num;
  	info->iommu	= iommu;
-	curr = xa_cmpxchg(&domain->iommu_array, iommu->seq_id,
-			  NULL, info, GFP_KERNEL);
-	if (curr) {
-		ret = xa_err(curr) ? : -EBUSY;
-		goto err_clear;
-	}
Quoted 14 lines

drivers/iommu/intel/iommu.c int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu) · line 1369

Quoted 10 lines
drivers/iommu/intel/iommu.c +2 -7
-	return 0;
-
-err_clear:
-	ida_free(&iommu->domain_ida, info->did);
-err_unlock:
-	kfree(info);
-	return ret;
+	return xa_err(xa_store(&domain->iommu_array, iommu->seq_id,
+			       no_free_ptr(info), GFP_KERNEL));
  }

no_free_ptr() should be used before successful return. Here xa_store() could return error but at that point no auto free as no_free_ptr() already changes 'info' to NULL. then memory leak.

Hmm, I've considered this. My thought was that xa_store() failure only occurs due to the system running out of memory, and the Linux kernel can't recover from it. In that case, the system is already broken; hence, handling the failure case here doesn't make things better.

Thanks,
baolu

#10 lore raw

Re: [PATCH 3/3] iommu/vt-d: Simplify domain_attach_iommu()

Quoted 13 lines · Baolu Lu
Quoted 4 lines
Quoted 4 lines
drivers/iommu/intel/iommu.c +0 -4
-err_clear:
-	ida_free(&iommu->domain_ida, info->did);
-err_unlock:
-	kfree(info);
Quoted 8 lines

drivers/iommu/intel/iommu.c int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu) · line 1369

Quoted 4 lines
drivers/iommu/intel/iommu.c +2 -1
-	return ret;
+	return xa_err(xa_store(&domain->iommu_array, iommu->seq_id,
+			       no_free_ptr(info), GFP_KERNEL));
  }

no_free_ptr() should be used before successful return. Here xa_store() could return error but at that point no auto free as no_free_ptr() already changes 'info' to NULL. then memory leak.

Quoted 4 lines

Hmm, I've considered this. My thought was that xa_store() failure only occurs due to the system running out of memory, and the Linux kernel can't recover from it. In that case, the system is already broken; hence, handling the failure case here doesn't make things better.

That's not the kernel pattern, you are supposed to unwind correctly in those failures

I think you should not use cleanup.h for something complicated like this..

Jason

author maintainer
#11 lore raw

Re: [PATCH 3/3] iommu/vt-d: Simplify domain_attach_iommu()

Quoted 17 lines · Jason Gunthorpe
Quoted 11 lines · Baolu Lu
Quoted 11 lines

drivers/iommu/intel/iommu.c int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu) · line 1369

Quoted 8 lines
drivers/iommu/intel/iommu.c +2 -5
-err_clear:
-	ida_free(&iommu->domain_ida, info->did);
-err_unlock:
-	kfree(info);
-	return ret;
+	return xa_err(xa_store(&domain->iommu_array, iommu->seq_id,
+			       no_free_ptr(info), GFP_KERNEL));
   }

no_free_ptr() should be used before successful return. Here xa_store() could return error but at that point no auto free as no_free_ptr() already changes 'info' to NULL. then memory leak.

Quoted 4 lines

Hmm, I've considered this. My thought was that xa_store() failure only occurs due to the system running out of memory, and the Linux kernel can't recover from it. In that case, the system is already broken; hence, handling the failure case here doesn't make things better.

Quoted 5 lines

That's not the kernel pattern, you are supposed to unwind correctly in those failures

I think you should not use cleanup.h for something complicated like this..

Okay, so let me drop this patch.

Thanks,
baolu

#12 lore raw

Re: [PATCH 3/3] iommu/vt-d: Simplify domain_attach_iommu()

In reply to Lu Baolu
Quoted 33 lines · Lu Baolu

Use the __free(kfree) attribute with kzalloc() to automatically handle the freeing of the allocated struct iommu_domain_info on error or early exit paths, eliminating the need for explicit kfree() calls in error handling branches.

---
 drivers/iommu/intel/iommu.c | 29 ++++++++---------------------
 1 file changed, 8 insertions(+), 21 deletions(-)
drivers/iommu/intel/iommu.c +4 -3
index 3a9ea0ad2cd3..12382c85495f 100644
@@ -1337,13 +1337,14 @@ static bool first_level_by_default(struct intel_iommu *iommu)
 
 int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
 {
-	struct iommu_domain_info *info, *curr;
-	int num, ret = -ENOSPC;
+	struct iommu_domain_info *curr;
+	int num;
 
 	if (domain->domain.type == IOMMU_DOMAIN_SVA)
 		return 0;
 
-	info = kzalloc(sizeof(*info), GFP_KERNEL);
+	struct iommu_domain_info *info __free(kfree) =
+		kzalloc(sizeof(*info), GFP_KERNEL);
 	if (!info)
 		return -ENOMEM;
 
@@ -1351,34 +1352,20 @@ int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)

[..]

drivers/iommu/intel/iommu.c int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu) · line 1369

Quoted 8 lines
drivers/iommu/intel/iommu.c +2 -5
-err_clear:
-	ida_free(&iommu->domain_ida, info->did);
-err_unlock:
-	kfree(info);
-	return ret;
+	return xa_err(xa_store(&domain->iommu_array, iommu->seq_id,
+			       no_free_ptr(info), GFP_KERNEL));
 }

This pattern looks like it wants a "xa_store_or_{reset,kfree}()" helper that handles both canceling a scope based cleanup and taking responsibility for error-exit-freeing @info in one statement.

I.e. this is similar to a:

	"return devm_add_action_or_reset(..., no_free_ptr(obj), ...)"

...pattern.

author maintainer
#13 lore raw

Re: [PATCH 3/3] iommu/vt-d: Simplify domain_attach_iommu()

Quoted 48 lines · Dan Williams
Quoted 33 lines · Lu Baolu

Use the __free(kfree) attribute with kzalloc() to automatically handle the freeing of the allocated struct iommu_domain_info on error or early exit paths, eliminating the need for explicit kfree() calls in error handling branches.

---
  drivers/iommu/intel/iommu.c | 29 ++++++++---------------------
  1 file changed, 8 insertions(+), 21 deletions(-)
drivers/iommu/intel/iommu.c +4 -3
index 3a9ea0ad2cd3..12382c85495f 100644
@@ -1337,13 +1337,14 @@ static bool first_level_by_default(struct intel_iommu *iommu)
 
  int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)
  {
-	struct iommu_domain_info *info, *curr;
-	int num, ret = -ENOSPC;
+	struct iommu_domain_info *curr;
+	int num;
 
  	if (domain->domain.type == IOMMU_DOMAIN_SVA)
  		return 0;
 
-	info = kzalloc(sizeof(*info), GFP_KERNEL);
+	struct iommu_domain_info *info __free(kfree) =
+		kzalloc(sizeof(*info), GFP_KERNEL);
  	if (!info)
  		return -ENOMEM;
 
@@ -1351,34 +1352,20 @@ int domain_attach_iommu(struct dmar_domain *domain, struct intel_iommu *iommu)

[..]

drivers/iommu/intel/iommu.c

Quoted 8 lines
drivers/iommu/intel/iommu.c +2 -5
-err_clear:
-	ida_free(&iommu->domain_ida, info->did);
-err_unlock:
-	kfree(info);
-	return ret;
+	return xa_err(xa_store(&domain->iommu_array, iommu->seq_id,
+			       no_free_ptr(info), GFP_KERNEL));
  }

This pattern looks like it wants a "xa_store_or_{reset,kfree}()" helper that handles both canceling a scope based cleanup and taking responsibility for error-exit-freeing @info in one statement.

Quoted 6 lines

I.e. this is similar to a:

	"return devm_add_action_or_reset(..., no_free_ptr(obj), ...)"

...pattern.

Yes. Perhaps adding a xa_store variant would be beneficial in all places that require this pattern.

Something like this?

include/linux/xarray.h +29 -0
index 78eede109b1a..efbdff7ebda4 100644
@@ -626,6 +626,35 @@ static inline void *xa_store_irq(struct xarray *xa, unsigned long index,
        return curr;
 }
 
+/**
+ * xa_store_or_kfree() - Store this entry in the XArray.
+ * @xa: XArray.
+ * @index: Index into array.
+ * @entry: New entry.
+ * @gfp: Memory allocation flags.
+ *
+ * This function is like calling xa_store() except it kfrees the new
+ * entry if an error happened.
+ *
+ * Context: Process context. Any context. Takes and releases the xa_lock.
+ * May sleep if the @gfp flags permit.
+ * Return: The old entry at this index or xa_err() if an error happened.
+ */
+static inline void *xa_store_or_kfree(struct xarray *xa, unsigned long index,
+               void *entry, gfp_t gfp)
+{
+       void *curr;
+
+       xa_lock(xa);
+       curr = __xa_store(xa, index, entry, gfp);
+       xa_unlock(xa);
+
+       if (xa_err(curr))
+               kfree(entry);
+
+       return curr;
+}
+
 /**
  * xa_erase_bh() - Erase this entry from the XArray.
  * @xa: XArray.

Thanks,
baolu