iommu/arm-smmu-v3: Add a threshold to avoid potential soft lockup

Message ID 20240115114040.6279-1-zhangzekun11@huawei.com
State New
Headers
Series iommu/arm-smmu-v3: Add a threshold to avoid potential soft lockup |

Commit Message

zhangzekun (A) Jan. 15, 2024, 11:40 a.m. UTC
  The commit d5afb4b47e13 ("iommu/arm-smmu-v3: Fix soft lockup triggered
by arm_smmu_mm_invalidate_range") has fix a soft lockup problem when
running SVA case, but code paths from iommu_unmap and dma APIs still
remain unfixed which could also cause potential soft lockup.

When cmdq is quite busy and don't have much space for batch submitting
cmds, and size passed to __arm_smmu_tlb_inv_range() is large (1G in this
case), the following softlockup is triggered.

WARN: soft lockup - CPU#71 stuck for 12s! [qemu-kvm:1303]
..
Call trace:
 dump_backtrace+0x0/0x200
 show_stack+0x20/0x30
 dump_stack+0xf0/0x138
 watchdog_print_info+0x48/0x54
 watchdog_process_before_softlockup+0x9c/0xa0
 watchdog_timer_fn+0x1ac/0x2f0
 __run_hrtimer+0x98/0x2b4
 __hrtimer_run_queues+0xc0/0x13c
 hrtimer_interrupt+0x150/0x3e4
 arch_timer_handler_phys+0x3c/0x50
 handle_percpu_devid_irq+0x90/0x1f4
 __handle_domain_irq+0x84/0xfc
 gic_handle_irq+0x88/0x2b0
 el1_irq+0xb8/0x140
 arm_smmu_cmdq_issue_cmdlist+0x184/0x5f4
 __arm_smmu_tlb_inv_range+0x114/0x22c
 arm_smmu_tlb_inv_walk+0x88/0x120
 __arm_lpae_unmap+0x188/0x2c0
 __arm_lpae_unmap+0x104/0x2c0
 arm_lpae_unmap+0x68/0x80
 arm_smmu_unmap+0x24/0x40
 __iommu_unmap+0xd8/0x210
 iommu_unmap+0x44/0x9c
..

The basic idea is use the actual granual size instead of PAGE_SIZE used
in SVA scenarios to calculate a threshold. When smmu without
ARM_SMMU_FEAT_RANGE_INV need to invalid a TLB range larger than the
threshold, we use the granularity of asid or vmid to invalid the TLB. The
calculation logic is similar to calculate 'bits_per_level' when allocating
io-pgtable, which could also been applyed to calculate the existing
threshold in SVA scenarios. Besides, change the comment "MAX_TLBI_OPS" to
"MAX_DVM_OPS", because it is has been renamed in commit ec1c3b9ff160
("arm64: tlbflush: Rename MAX_TLBI_OPS")

Signed-off-by: Zhang Zekun <zhangzekun11@huawei.com>
---
 .../iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c   | 11 +--------
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c   | 23 +++++++++++++++----
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h   | 10 ++++++++
 3 files changed, 30 insertions(+), 14 deletions(-)
  

Comments

Jason Gunthorpe Jan. 15, 2024, 3:31 p.m. UTC | #1
On Mon, Jan 15, 2024 at 07:40:40PM +0800, Zhang Zekun wrote:
> The commit d5afb4b47e13 ("iommu/arm-smmu-v3: Fix soft lockup triggered
> by arm_smmu_mm_invalidate_range") has fix a soft lockup problem when
> running SVA case, but code paths from iommu_unmap and dma APIs still
> remain unfixed which could also cause potential soft lockup.
> 
> When cmdq is quite busy and don't have much space for batch submitting
> cmds, and size passed to __arm_smmu_tlb_inv_range() is large (1G in this
> case), the following softlockup is triggered.
> 
> WARN: soft lockup - CPU#71 stuck for 12s! [qemu-kvm:1303]
> ...
> Call trace:
>  dump_backtrace+0x0/0x200
>  show_stack+0x20/0x30
>  dump_stack+0xf0/0x138
>  watchdog_print_info+0x48/0x54
>  watchdog_process_before_softlockup+0x9c/0xa0
>  watchdog_timer_fn+0x1ac/0x2f0
>  __run_hrtimer+0x98/0x2b4
>  __hrtimer_run_queues+0xc0/0x13c
>  hrtimer_interrupt+0x150/0x3e4
>  arch_timer_handler_phys+0x3c/0x50
>  handle_percpu_devid_irq+0x90/0x1f4
>  __handle_domain_irq+0x84/0xfc
>  gic_handle_irq+0x88/0x2b0
>  el1_irq+0xb8/0x140
>  arm_smmu_cmdq_issue_cmdlist+0x184/0x5f4
>  __arm_smmu_tlb_inv_range+0x114/0x22c
>  arm_smmu_tlb_inv_walk+0x88/0x120
>  __arm_lpae_unmap+0x188/0x2c0
>  __arm_lpae_unmap+0x104/0x2c0
>  arm_lpae_unmap+0x68/0x80
>  arm_smmu_unmap+0x24/0x40
>  __iommu_unmap+0xd8/0x210
>  iommu_unmap+0x44/0x9c
> ...

What is the rest of the call chain? How did you get into such a large
invalidation?

> @@ -228,7 +219,7 @@ static void arm_smmu_mm_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
>  	 */
>  	size = end - start;
>  	if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_RANGE_INV)) {
> -		if (size >= CMDQ_MAX_TLBI_OPS * PAGE_SIZE)
> +		if (size >= CMDQ_MAX_TLBI_OPS(PAGE_SIZE) * PAGE_SIZE)
>  			size = 0;
>  	} else {
>  		if (size == ULONG_MAX)

I would like to see the SVA code rely on the common invalidation
infrastructure, this should be pushed down into the invalidation logic
not be in the SVA code. With some adjustments we can get common
functions for all this and push the decision making into the actual
function.

Something like this, on top of my "part 3" branch
https://github.com/jgunthorpe/linux/commits/smmuv3_newapi/
split into more patches:

diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
index 5ab976f6b108dd..b30cce97b01d06 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
@@ -30,7 +30,6 @@ static int arm_smmu_realloc_s1_domain_asid(struct arm_smmu_device *smmu,
 					   struct arm_smmu_domain *smmu_domain)
 {
 	struct arm_smmu_master_domain *master_domain;
-	u32 old_asid = smmu_domain->asid;
 	struct arm_smmu_cd target_cd;
 	unsigned long flags;
 	int ret;
@@ -68,9 +67,6 @@ static int arm_smmu_realloc_s1_domain_asid(struct arm_smmu_device *smmu,
 					&target_cd);
 	}
 	spin_unlock_irqrestore(&smmu_domain->devices_lock, flags);
-
-	/* Clean the ASID we are about to assign to a new translation */
-	arm_smmu_tlb_inv_asid(smmu, old_asid);
 	return 0;
 }
 
@@ -175,24 +171,12 @@ static void arm_smmu_mm_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
 	 * range. So do a simple translation here by calculating size correctly.
 	 */
 	size = end - start;
-	if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_RANGE_INV)) {
-		if (size >= CMDQ_MAX_TLBI_OPS * PAGE_SIZE)
-			size = 0;
-	} else {
-		if (size == ULONG_MAX)
-			size = 0;
-	}
+	if (size == ULONG_MAX)
+		size = 0;
 
-	if (!smmu_domain->btm_invalidation) {
-		ioasid_t asid = READ_ONCE(smmu_domain->asid);
-
-		if (!size)
-			arm_smmu_tlb_inv_asid(smmu_domain->smmu, asid);
-		else
-			arm_smmu_tlb_inv_range_asid(start, size, asid,
-						    PAGE_SIZE, false,
-						    smmu_domain);
-	}
+	if (!smmu_domain->btm_invalidation)
+		arm_smmu_tlb_inv_range_s1(start, size, PAGE_SIZE, false,
+					  smmu_domain);
 
 	arm_smmu_atc_inv_domain(smmu_domain, start, size);
 }
@@ -228,7 +212,7 @@ static void arm_smmu_mm_release(struct mmu_notifier *mn, struct mm_struct *mm)
 	}
 	spin_unlock_irqrestore(&smmu_domain->devices_lock, flags);
 
-	arm_smmu_tlb_inv_asid(smmu_domain->smmu, smmu_domain->asid);
+	arm_smmu_tlb_inv_all_s1(smmu_domain);
 	arm_smmu_atc_inv_domain(smmu_domain, 0, 0);
 }
 
@@ -506,6 +490,8 @@ static int arm_smmu_share_asid(struct arm_smmu_device *smmu,
 		ret = arm_smmu_realloc_s1_domain_asid(smmu, old_s1_domain);
 		if (ret)
 			goto out_restore_s1;
+		/* Clean the ASID since it was just recovered */
+		arm_smmu_tlb_inv_all_s1(smmu_domain);
 	}
 
 	smmu_domain->btm_invalidation = true;
diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
index b07e0692eb78ea..79c85791b0c3d5 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
@@ -90,6 +90,7 @@ static void arm_smmu_rmr_install_bypass_ste(struct arm_smmu_device *smmu);
 static int arm_smmu_domain_finalise(struct arm_smmu_domain *smmu_domain,
 				    struct arm_smmu_device *smmu);
 static int arm_smmu_alloc_cd_tables(struct arm_smmu_master *master);
+static void arm_smmu_tlb_inv_all_s2(struct arm_smmu_domain *smmu_domain);
 
 static void parse_driver_options(struct arm_smmu_device *smmu)
 {
@@ -965,17 +966,6 @@ static int arm_smmu_page_response(struct device *dev,
 }
 
 /* Context descriptor manipulation functions */
-void arm_smmu_tlb_inv_asid(struct arm_smmu_device *smmu, u16 asid)
-{
-	struct arm_smmu_cmdq_ent cmd = {
-		.opcode	= smmu->features & ARM_SMMU_FEAT_E2H ?
-			CMDQ_OP_TLBI_EL2_ASID : CMDQ_OP_TLBI_NH_ASID,
-		.tlbi.asid = asid,
-	};
-
-	arm_smmu_cmdq_issue_cmd_with_sync(smmu, &cmd);
-}
-
 #define NUM_ENTRY_QWORDS                                                \
 	(max(sizeof(struct arm_smmu_ste), sizeof(struct arm_smmu_cd)) / \
 	 sizeof(u64))
@@ -2106,8 +2096,6 @@ int arm_smmu_atc_inv_domain(struct arm_smmu_domain *smmu_domain,
 static void arm_smmu_tlb_inv_context(void *cookie)
 {
 	struct arm_smmu_domain *smmu_domain = cookie;
-	struct arm_smmu_device *smmu = smmu_domain->smmu;
-	struct arm_smmu_cmdq_ent cmd;
 
 	/*
 	 * NOTE: when io-pgtable is in non-strict mode, we may get here with
@@ -2116,13 +2104,10 @@ static void arm_smmu_tlb_inv_context(void *cookie)
 	 * insertion to guarantee those are observed before the TLBI. Do be
 	 * careful, 007.
 	 */
-	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1) {
-		arm_smmu_tlb_inv_asid(smmu, READ_ONCE(smmu_domain->asid));
-	} else {
-		cmd.opcode	= CMDQ_OP_TLBI_S12_VMALL;
-		cmd.tlbi.vmid	= smmu_domain->vmid;
-		arm_smmu_cmdq_issue_cmd_with_sync(smmu, &cmd);
-	}
+	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1)
+		arm_smmu_tlb_inv_all_s1(smmu_domain);
+	else
+		arm_smmu_tlb_inv_all_s2(smmu_domain);
 	arm_smmu_atc_inv_domain(smmu_domain, 0, 0);
 }
 
@@ -2197,47 +2182,69 @@ static void __arm_smmu_tlb_inv_range(struct arm_smmu_cmdq_ent *cmd,
 	arm_smmu_cmdq_batch_submit(smmu, &cmds);
 }
 
-static void arm_smmu_tlb_inv_range_domain(unsigned long iova, size_t size,
-					  size_t granule, bool leaf,
-					  struct arm_smmu_domain *smmu_domain)
+static bool arm_smmu_inv_range_too_big(struct arm_smmu_device *smmu,
+				       size_t size, size_t granule)
 {
-	struct arm_smmu_cmdq_ent cmd = {
-		.tlbi = {
-			.leaf	= leaf,
-		},
-	};
+	unsigned max_ops;
 
-	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1) {
-		cmd.opcode	= smmu_domain->smmu->features & ARM_SMMU_FEAT_E2H ?
-				  CMDQ_OP_TLBI_EL2_VA : CMDQ_OP_TLBI_NH_VA;
-		cmd.tlbi.asid	= smmu_domain->asid;
-	} else {
-		cmd.opcode	= CMDQ_OP_TLBI_S2_IPA;
-		cmd.tlbi.vmid	= smmu_domain->vmid;
-	}
-	__arm_smmu_tlb_inv_range(&cmd, iova, size, granule, smmu_domain);
+	/* 0 size means invalidate all */
+	if (!size || size == SIZE_MAX)
+		return true;
 
-	/*
-	 * Unfortunately, this can't be leaf-only since we may have
-	 * zapped an entire table.
-	 */
-	arm_smmu_atc_inv_domain(smmu_domain, iova, size);
+	if (smmu->features & ARM_SMMU_FEAT_RANGE_INV)
+		return false;
+
+	max_ops = 1 << (ilog2(granule) - 3);
+	return size >= max_ops * granule;
 }
 
-void arm_smmu_tlb_inv_range_asid(unsigned long iova, size_t size, int asid,
-				 size_t granule, bool leaf,
-				 struct arm_smmu_domain *smmu_domain)
+void arm_smmu_tlb_inv_range_s1(unsigned long iova, size_t size, size_t granule,
+			       bool leaf, struct arm_smmu_domain *smmu_domain)
 {
 	struct arm_smmu_cmdq_ent cmd = {
 		.opcode	= smmu_domain->smmu->features & ARM_SMMU_FEAT_E2H ?
 			  CMDQ_OP_TLBI_EL2_VA : CMDQ_OP_TLBI_NH_VA,
 		.tlbi = {
-			.asid	= asid,
+			.asid	= READ_ONCE(smmu_domain->asid),
 			.leaf	= leaf,
 		},
 	};
 
-	__arm_smmu_tlb_inv_range(&cmd, iova, size, granule, smmu_domain);
+	if (arm_smmu_inv_range_too_big(smmu_domain->smmu, size, granule)) {
+		cmd.opcode = smmu_domain->smmu->features & ARM_SMMU_FEAT_E2H ?
+				     CMDQ_OP_TLBI_EL2_ASID :
+				     CMDQ_OP_TLBI_NH_ASID,
+		arm_smmu_cmdq_issue_cmd_with_sync(smmu_domain->smmu, &cmd);
+	} else{
+		__arm_smmu_tlb_inv_range(&cmd, iova, size, granule,
+					 smmu_domain);
+	}
+}
+
+static void arm_smmu_tlb_inv_range_s2(unsigned long iova, size_t size,
+				      size_t granule, bool leaf,
+				      struct arm_smmu_domain *smmu_domain)
+{
+	struct arm_smmu_cmdq_ent cmd = {
+		.opcode	= CMDQ_OP_TLBI_S2_IPA,
+		.tlbi = {
+			.vmid	= smmu_domain->vmid,
+			.leaf	= leaf,
+		},
+	};
+
+	if (arm_smmu_inv_range_too_big(smmu_domain->smmu, size, granule)) {
+		cmd.opcode = CMDQ_OP_TLBI_S12_VMALL;
+		arm_smmu_cmdq_issue_cmd_with_sync(smmu_domain->smmu, &cmd);
+	} else {
+		__arm_smmu_tlb_inv_range(&cmd, iova, size, granule,
+					 smmu_domain);
+	}
+}
+
+static void arm_smmu_tlb_inv_all_s2(struct arm_smmu_domain *smmu_domain)
+{
+	arm_smmu_tlb_inv_range_s2(0, 0, PAGE_SIZE, false, smmu_domain);
 }
 
 static void arm_smmu_tlb_inv_page_nosync(struct iommu_iotlb_gather *gather,
@@ -2253,7 +2260,15 @@ static void arm_smmu_tlb_inv_page_nosync(struct iommu_iotlb_gather *gather,
 static void arm_smmu_tlb_inv_walk(unsigned long iova, size_t size,
 				  size_t granule, void *cookie)
 {
-	arm_smmu_tlb_inv_range_domain(iova, size, granule, false, cookie);
+	struct arm_smmu_domain *smmu_domain = cookie;
+
+	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1)
+		arm_smmu_tlb_inv_range_s1(iova, size, granule, false,
+					  smmu_domain);
+	else
+		arm_smmu_tlb_inv_range_s2(iova, size, granule, false,
+					  smmu_domain);
+	arm_smmu_atc_inv_domain(smmu_domain, iova, size);
 }
 
 static const struct iommu_flush_ops arm_smmu_flush_ops = {
@@ -2379,7 +2394,7 @@ void arm_smmu_domain_free_id(struct arm_smmu_domain *smmu_domain)
 	if ((smmu_domain->stage == ARM_SMMU_DOMAIN_S1 ||
 	     smmu_domain->domain.type == IOMMU_DOMAIN_SVA) &&
 	    smmu_domain->asid) {
-		arm_smmu_tlb_inv_asid(smmu, smmu_domain->asid);
+		arm_smmu_tlb_inv_all_s1(smmu_domain);
 
 		/* Prevent SVA from touching the CD while we're freeing it */
 		mutex_lock(&smmu->asid_lock);
@@ -3183,13 +3198,23 @@ static void arm_smmu_iotlb_sync(struct iommu_domain *domain,
 				struct iommu_iotlb_gather *gather)
 {
 	struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain);
+	size_t size = gather->end - gather->start + 1;
 
 	if (!gather->pgsize)
 		return;
 
-	arm_smmu_tlb_inv_range_domain(gather->start,
-				      gather->end - gather->start + 1,
-				      gather->pgsize, true, smmu_domain);
+	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1)
+		arm_smmu_tlb_inv_range_s1(gather->start, size, gather->pgsize,
+					  true, smmu_domain);
+	else
+		arm_smmu_tlb_inv_range_s2(gather->start, size, gather->pgsize,
+					  true, smmu_domain);
+
+	/*
+	 * Unfortunately, this can't be leaf-only since we may have
+	 * zapped an entire table.
+	 */
+	arm_smmu_atc_inv_domain(smmu_domain, gather->start, size);
 }
 
 static phys_addr_t
diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
index d65c716d4fc04f..9f5537ef97caff 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
@@ -834,10 +834,13 @@ void arm_smmu_remove_pasid(struct arm_smmu_master *master,
 int arm_smmu_domain_alloc_id(struct arm_smmu_device *smmu,
 			     struct arm_smmu_domain *smmu_domain);
 void arm_smmu_domain_free_id(struct arm_smmu_domain *smmu_domain);
-void arm_smmu_tlb_inv_asid(struct arm_smmu_device *smmu, u16 asid);
-void arm_smmu_tlb_inv_range_asid(unsigned long iova, size_t size, int asid,
-				 size_t granule, bool leaf,
-				 struct arm_smmu_domain *smmu_domain);
+void arm_smmu_tlb_inv_range_s1(unsigned long iova, size_t size, size_t granule,
+			       bool leaf, struct arm_smmu_domain *smmu_domain);
+static inline void arm_smmu_tlb_inv_all_s1(struct arm_smmu_domain *smmu_domain)
+{
+	arm_smmu_tlb_inv_range_s1(0, 0, PAGE_SIZE, false, smmu_domain);
+}
+
 int arm_smmu_atc_inv_domain(struct arm_smmu_domain *smmu_domain,
 			    unsigned long iova, size_t size);
  
zhangzekun (A) Jan. 18, 2024, 11:21 a.m. UTC | #2
在 2024/1/15 23:31, Jason Gunthorpe 写道:
> On Mon, Jan 15, 2024 at 07:40:40PM +0800, Zhang Zekun wrote:
>> The commit d5afb4b47e13 ("iommu/arm-smmu-v3: Fix soft lockup triggered
>> by arm_smmu_mm_invalidate_range") has fix a soft lockup problem when
>> running SVA case, but code paths from iommu_unmap and dma APIs still
>> remain unfixed which could also cause potential soft lockup.
>>
>> When cmdq is quite busy and don't have much space for batch submitting
>> cmds, and size passed to __arm_smmu_tlb_inv_range() is large (1G in this
>> case), the following softlockup is triggered.
>>
>> WARN: soft lockup - CPU#71 stuck for 12s! [qemu-kvm:1303]
>> ...
>> Call trace:
>>   dump_backtrace+0x0/0x200
>>   show_stack+0x20/0x30
>>   dump_stack+0xf0/0x138
>>   watchdog_print_info+0x48/0x54
>>   watchdog_process_before_softlockup+0x9c/0xa0
>>   watchdog_timer_fn+0x1ac/0x2f0
>>   __run_hrtimer+0x98/0x2b4
>>   __hrtimer_run_queues+0xc0/0x13c
>>   hrtimer_interrupt+0x150/0x3e4
>>   arch_timer_handler_phys+0x3c/0x50
>>   handle_percpu_devid_irq+0x90/0x1f4
>>   __handle_domain_irq+0x84/0xfc
>>   gic_handle_irq+0x88/0x2b0
>>   el1_irq+0xb8/0x140
>>   arm_smmu_cmdq_issue_cmdlist+0x184/0x5f4
>>   __arm_smmu_tlb_inv_range+0x114/0x22c
>>   arm_smmu_tlb_inv_walk+0x88/0x120
>>   __arm_lpae_unmap+0x188/0x2c0
>>   __arm_lpae_unmap+0x104/0x2c0
>>   arm_lpae_unmap+0x68/0x80
>>   arm_smmu_unmap+0x24/0x40
>>   __iommu_unmap+0xd8/0x210
>>   iommu_unmap+0x44/0x9c
>> ...
> What is the rest of the call chain? How did you get into such a large
> invalidation?
we are doing some test with qemu virtual machines. The testing platform 
has 96 cores and 960GB memories. It will enable 64 virtual machines and 
every virtual machine will cost 8G memories. When batching closing these 
64 virtaul machine concurrently, we will get soft lockup.

  WARN: soft lockup - CPU#18 stuck for 11s! [worker:3219]
...
Call trace:
  dump_backtrace+0x0/0x200
  show_stack+0x20/0x30
  dump_stack+0xf0/0x138
  watchdog_print_info+0x48/0x54
  watchdog_process_before_softlockup+0x9c/0xa0
  watchdog_timer_fn+0x1ac/0x2f0
  __run_hrtimer+0x98/0x2b4
  __hrtimer_run_queues+0xc0/0x13c
  hrtimer_interrupt+0x150/0x3e4
  arch_timer_handler_phys+0x3c/0x50
  handle_percpu_devid_irq+0x90/0x1f4
  __handle_domain_irq+0x84/0xfc
  gic_handle_irq+0x88/0x2b0
  el1_irq+0xb8/0x140
  ktime_get+0x3c/0xac
  arm_smmu_cmdq_poll_until_not_full+0xb0/0x1a0
  arm_smmu_cmdq_issue_cmdlist+0x194/0x5f4
  __arm_smmu_tlb_inv_range+0x114/0x22c
  arm_smmu_tlb_inv_walk+0x88/0x120
  __arm_lpae_unmap+0x188/0x2c0
  __arm_lpae_unmap+0x104/0x2c0
  arm_lpae_unmap+0x68/0x80
  arm_smmu_unmap+0x24/0x40
  __iommu_unmap+0xd8/0x210
  iommu_unmap_fast+0x18/0x24
  unmap_unpin_fast+0x7c/0x140 [vfio_iommu_type1]
  vfio_unmap_unpin+0x14c/0x3ac [vfio_iommu_type1]
  vfio_remove_dma+0x38/0x124 [vfio_iommu_type1]
  vfio_iommu_type1_detach_group+0x4b8/0x4e0 [vfio_iommu_type1]
  __vfio_group_unset_container+0x58/0x18c [vfio]
  vfio_group_try_dissolve_container+0x80/0x94 [vfio]
  vfio_group_put_external_user+0x20/0x54 [vfio]
  kvm_vfio_destroy+0xa8/0x12c
  kvm_destroy_vm+0x20c/0x300
  kvm_put_kvm+0x74/0xa0
  kvm_vcpu_release+0x20/0x30
  __fput+0xc4/0x270
  ____fput+0x18/0x24
  task_work_run+0xd0/0x1a0
  do_exit+0x1d8/0x480
  do_group_exit+0x40/0x130
  get_signal+0x1f8/0x744
  do_signal+0x98/0x204
  do_notify_resume+0x15c/0x1e0
  work_pending+0xc/0xa4
>> @@ -228,7 +219,7 @@ static void arm_smmu_mm_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
>>   	 */
>>   	size = end - start;
>>   	if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_RANGE_INV)) {
>> -		if (size >= CMDQ_MAX_TLBI_OPS * PAGE_SIZE)
>> +		if (size >= CMDQ_MAX_TLBI_OPS(PAGE_SIZE) * PAGE_SIZE)
>>   			size = 0;
>>   	} else {
>>   		if (size == ULONG_MAX)
> I would like to see the SVA code rely on the common invalidation
> infrastructure, this should be pushed down into the invalidation logic
> not be in the SVA code. With some adjustments we can get common
> functions for all this and push the decision making into the actual
> function.
Thanks for your patch, the patch described below make sense to me. After 
spliting TLB invalidate operations according to domain stage, 
arm_smmu_tlb_inv_*_s1() can be used both in SVA codes and original smmu 
codes, and can implement arm_smmu_tlb_inv_all_s1() just using 
arm_smmu_tlb_inv_range_s1() by passing a special size. But the current 
smmu-v3 driver has not move "asid" into "struct smmu_domain", we still 
need to pass the exact asid from SVA.
>
> Something like this, on top of my "part 3" branch
> https://github.com/jgunthorpe/linux/commits/smmuv3_newapi/
> split into more patches:
>
> diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
> index 5ab976f6b108dd..b30cce97b01d06 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
> +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
> @@ -30,7 +30,6 @@ static int arm_smmu_realloc_s1_domain_asid(struct arm_smmu_device *smmu,
>   					   struct arm_smmu_domain *smmu_domain)
>   {
>   	struct arm_smmu_master_domain *master_domain;
> -	u32 old_asid = smmu_domain->asid;
>   	struct arm_smmu_cd target_cd;
>   	unsigned long flags;
>   	int ret;
> @@ -68,9 +67,6 @@ static int arm_smmu_realloc_s1_domain_asid(struct arm_smmu_device *smmu,
>   					&target_cd);
>   	}
>   	spin_unlock_irqrestore(&smmu_domain->devices_lock, flags);
> -
> -	/* Clean the ASID we are about to assign to a new translation */
> -	arm_smmu_tlb_inv_asid(smmu, old_asid);
>   	return 0;
>   }
>   
> @@ -175,24 +171,12 @@ static void arm_smmu_mm_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
>   	 * range. So do a simple translation here by calculating size correctly.
>   	 */
>   	size = end - start;
> -	if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_RANGE_INV)) {
> -		if (size >= CMDQ_MAX_TLBI_OPS * PAGE_SIZE)
> -			size = 0;
> -	} else {
> -		if (size == ULONG_MAX)
> -			size = 0;
> -	}
> +	if (size == ULONG_MAX)
> +		size = 0;
>   
> -	if (!smmu_domain->btm_invalidation) {
> -		ioasid_t asid = READ_ONCE(smmu_domain->asid);
> -
> -		if (!size)
> -			arm_smmu_tlb_inv_asid(smmu_domain->smmu, asid);
> -		else
> -			arm_smmu_tlb_inv_range_asid(start, size, asid,
> -						    PAGE_SIZE, false,
> -						    smmu_domain);
> -	}
> +	if (!smmu_domain->btm_invalidation)
> +		arm_smmu_tlb_inv_range_s1(start, size, PAGE_SIZE, false,
> +					  smmu_domain);
>   
>   	arm_smmu_atc_inv_domain(smmu_domain, start, size);
>   }
> @@ -228,7 +212,7 @@ static void arm_smmu_mm_release(struct mmu_notifier *mn, struct mm_struct *mm)
>   	}
>   	spin_unlock_irqrestore(&smmu_domain->devices_lock, flags);
>   
> -	arm_smmu_tlb_inv_asid(smmu_domain->smmu, smmu_domain->asid);
> +	arm_smmu_tlb_inv_all_s1(smmu_domain);
>   	arm_smmu_atc_inv_domain(smmu_domain, 0, 0);
>   }
>   
> @@ -506,6 +490,8 @@ static int arm_smmu_share_asid(struct arm_smmu_device *smmu,
>   		ret = arm_smmu_realloc_s1_domain_asid(smmu, old_s1_domain);
>   		if (ret)
>   			goto out_restore_s1;
> +		/* Clean the ASID since it was just recovered */
> +		arm_smmu_tlb_inv_all_s1(smmu_domain);
>   	}
>   
>   	smmu_domain->btm_invalidation = true;
> diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> index b07e0692eb78ea..79c85791b0c3d5 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> @@ -90,6 +90,7 @@ static void arm_smmu_rmr_install_bypass_ste(struct arm_smmu_device *smmu);
>   static int arm_smmu_domain_finalise(struct arm_smmu_domain *smmu_domain,
>   				    struct arm_smmu_device *smmu);
>   static int arm_smmu_alloc_cd_tables(struct arm_smmu_master *master);
> +static void arm_smmu_tlb_inv_all_s2(struct arm_smmu_domain *smmu_domain);
>   
>   static void parse_driver_options(struct arm_smmu_device *smmu)
>   {
> @@ -965,17 +966,6 @@ static int arm_smmu_page_response(struct device *dev,
>   }
>   
>   /* Context descriptor manipulation functions */
> -void arm_smmu_tlb_inv_asid(struct arm_smmu_device *smmu, u16 asid)
> -{
> -	struct arm_smmu_cmdq_ent cmd = {
> -		.opcode	= smmu->features & ARM_SMMU_FEAT_E2H ?
> -			CMDQ_OP_TLBI_EL2_ASID : CMDQ_OP_TLBI_NH_ASID,
> -		.tlbi.asid = asid,
> -	};
> -
> -	arm_smmu_cmdq_issue_cmd_with_sync(smmu, &cmd);
> -}
> -
>   #define NUM_ENTRY_QWORDS                                                \
>   	(max(sizeof(struct arm_smmu_ste), sizeof(struct arm_smmu_cd)) / \
>   	 sizeof(u64))
> @@ -2106,8 +2096,6 @@ int arm_smmu_atc_inv_domain(struct arm_smmu_domain *smmu_domain,
>   static void arm_smmu_tlb_inv_context(void *cookie)
>   {
>   	struct arm_smmu_domain *smmu_domain = cookie;
> -	struct arm_smmu_device *smmu = smmu_domain->smmu;
> -	struct arm_smmu_cmdq_ent cmd;
>   
>   	/*
>   	 * NOTE: when io-pgtable is in non-strict mode, we may get here with
> @@ -2116,13 +2104,10 @@ static void arm_smmu_tlb_inv_context(void *cookie)
>   	 * insertion to guarantee those are observed before the TLBI. Do be
>   	 * careful, 007.
>   	 */
> -	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1) {
> -		arm_smmu_tlb_inv_asid(smmu, READ_ONCE(smmu_domain->asid));
> -	} else {
> -		cmd.opcode	= CMDQ_OP_TLBI_S12_VMALL;
> -		cmd.tlbi.vmid	= smmu_domain->vmid;
> -		arm_smmu_cmdq_issue_cmd_with_sync(smmu, &cmd);
> -	}
> +	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1)
> +		arm_smmu_tlb_inv_all_s1(smmu_domain);
> +	else
> +		arm_smmu_tlb_inv_all_s2(smmu_domain);
>   	arm_smmu_atc_inv_domain(smmu_domain, 0, 0);
>   }
>   
> @@ -2197,47 +2182,69 @@ static void __arm_smmu_tlb_inv_range(struct arm_smmu_cmdq_ent *cmd,
>   	arm_smmu_cmdq_batch_submit(smmu, &cmds);
>   }
>   
> -static void arm_smmu_tlb_inv_range_domain(unsigned long iova, size_t size,
> -					  size_t granule, bool leaf,
> -					  struct arm_smmu_domain *smmu_domain)
> +static bool arm_smmu_inv_range_too_big(struct arm_smmu_device *smmu,
> +				       size_t size, size_t granule)
>   {
> -	struct arm_smmu_cmdq_ent cmd = {
> -		.tlbi = {
> -			.leaf	= leaf,
> -		},
> -	};
> +	unsigned max_ops;
>   
> -	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1) {
> -		cmd.opcode	= smmu_domain->smmu->features & ARM_SMMU_FEAT_E2H ?
> -				  CMDQ_OP_TLBI_EL2_VA : CMDQ_OP_TLBI_NH_VA;
> -		cmd.tlbi.asid	= smmu_domain->asid;
> -	} else {
> -		cmd.opcode	= CMDQ_OP_TLBI_S2_IPA;
> -		cmd.tlbi.vmid	= smmu_domain->vmid;
> -	}
> -	__arm_smmu_tlb_inv_range(&cmd, iova, size, granule, smmu_domain);
> +	/* 0 size means invalidate all */
> +	if (!size || size == SIZE_MAX)
> +		return true;
>   
> -	/*
> -	 * Unfortunately, this can't be leaf-only since we may have
> -	 * zapped an entire table.
> -	 */
> -	arm_smmu_atc_inv_domain(smmu_domain, iova, size);
> +	if (smmu->features & ARM_SMMU_FEAT_RANGE_INV)
> +		return false;
> +
> +	max_ops = 1 << (ilog2(granule) - 3);
> +	return size >= max_ops * granule;
>   }
>   
> -void arm_smmu_tlb_inv_range_asid(unsigned long iova, size_t size, int asid,
> -				 size_t granule, bool leaf,
> -				 struct arm_smmu_domain *smmu_domain)
> +void arm_smmu_tlb_inv_range_s1(unsigned long iova, size_t size, size_t granule,
> +			       bool leaf, struct arm_smmu_domain *smmu_domain)
>   {
>   	struct arm_smmu_cmdq_ent cmd = {
>   		.opcode	= smmu_domain->smmu->features & ARM_SMMU_FEAT_E2H ?
>   			  CMDQ_OP_TLBI_EL2_VA : CMDQ_OP_TLBI_NH_VA,
>   		.tlbi = {
> -			.asid	= asid,
> +			.asid	= READ_ONCE(smmu_domain->asid),
>   			.leaf	= leaf,
>   		},
>   	};
>   
> -	__arm_smmu_tlb_inv_range(&cmd, iova, size, granule, smmu_domain);
> +	if (arm_smmu_inv_range_too_big(smmu_domain->smmu, size, granule)) {
> +		cmd.opcode = smmu_domain->smmu->features & ARM_SMMU_FEAT_E2H ?
> +				     CMDQ_OP_TLBI_EL2_ASID :
> +				     CMDQ_OP_TLBI_NH_ASID,
> +		arm_smmu_cmdq_issue_cmd_with_sync(smmu_domain->smmu, &cmd);
> +	} else{
> +		__arm_smmu_tlb_inv_range(&cmd, iova, size, granule,
> +					 smmu_domain);
> +	}
> +}
> +
> +static void arm_smmu_tlb_inv_range_s2(unsigned long iova, size_t size,
> +				      size_t granule, bool leaf,
> +				      struct arm_smmu_domain *smmu_domain)
> +{
> +	struct arm_smmu_cmdq_ent cmd = {
> +		.opcode	= CMDQ_OP_TLBI_S2_IPA,
> +		.tlbi = {
> +			.vmid	= smmu_domain->vmid,
> +			.leaf	= leaf,
> +		},
> +	};
> +
> +	if (arm_smmu_inv_range_too_big(smmu_domain->smmu, size, granule)) {
> +		cmd.opcode = CMDQ_OP_TLBI_S12_VMALL;
> +		arm_smmu_cmdq_issue_cmd_with_sync(smmu_domain->smmu, &cmd);
> +	} else {
> +		__arm_smmu_tlb_inv_range(&cmd, iova, size, granule,
> +					 smmu_domain);
> +	}
> +}
> +
> +static void arm_smmu_tlb_inv_all_s2(struct arm_smmu_domain *smmu_domain)
> +{
> +	arm_smmu_tlb_inv_range_s2(0, 0, PAGE_SIZE, false, smmu_domain);
>   }
>   
>   static void arm_smmu_tlb_inv_page_nosync(struct iommu_iotlb_gather *gather,
> @@ -2253,7 +2260,15 @@ static void arm_smmu_tlb_inv_page_nosync(struct iommu_iotlb_gather *gather,
>   static void arm_smmu_tlb_inv_walk(unsigned long iova, size_t size,
>   				  size_t granule, void *cookie)
>   {
> -	arm_smmu_tlb_inv_range_domain(iova, size, granule, false, cookie);
> +	struct arm_smmu_domain *smmu_domain = cookie;
> +
> +	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1)
> +		arm_smmu_tlb_inv_range_s1(iova, size, granule, false,
> +					  smmu_domain);
> +	else
> +		arm_smmu_tlb_inv_range_s2(iova, size, granule, false,
> +					  smmu_domain);
> +	arm_smmu_atc_inv_domain(smmu_domain, iova, size);
>   }
>   
>   static const struct iommu_flush_ops arm_smmu_flush_ops = {
> @@ -2379,7 +2394,7 @@ void arm_smmu_domain_free_id(struct arm_smmu_domain *smmu_domain)
>   	if ((smmu_domain->stage == ARM_SMMU_DOMAIN_S1 ||
>   	     smmu_domain->domain.type == IOMMU_DOMAIN_SVA) &&
>   	    smmu_domain->asid) {
> -		arm_smmu_tlb_inv_asid(smmu, smmu_domain->asid);
> +		arm_smmu_tlb_inv_all_s1(smmu_domain);
>   
>   		/* Prevent SVA from touching the CD while we're freeing it */
>   		mutex_lock(&smmu->asid_lock);
> @@ -3183,13 +3198,23 @@ static void arm_smmu_iotlb_sync(struct iommu_domain *domain,
>   				struct iommu_iotlb_gather *gather)
>   {
>   	struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain);
> +	size_t size = gather->end - gather->start + 1;
>   
>   	if (!gather->pgsize)
>   		return;
>   
> -	arm_smmu_tlb_inv_range_domain(gather->start,
> -				      gather->end - gather->start + 1,
> -				      gather->pgsize, true, smmu_domain);
> +	if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1)
> +		arm_smmu_tlb_inv_range_s1(gather->start, size, gather->pgsize,
> +					  true, smmu_domain);
> +	else
> +		arm_smmu_tlb_inv_range_s2(gather->start, size, gather->pgsize,
> +					  true, smmu_domain);
> +
> +	/*
> +	 * Unfortunately, this can't be leaf-only since we may have
> +	 * zapped an entire table.
> +	 */
> +	arm_smmu_atc_inv_domain(smmu_domain, gather->start, size);
>   }
>   
>   static phys_addr_t
> diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
> index d65c716d4fc04f..9f5537ef97caff 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
> +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
> @@ -834,10 +834,13 @@ void arm_smmu_remove_pasid(struct arm_smmu_master *master,
>   int arm_smmu_domain_alloc_id(struct arm_smmu_device *smmu,
>   			     struct arm_smmu_domain *smmu_domain);
>   void arm_smmu_domain_free_id(struct arm_smmu_domain *smmu_domain);
> -void arm_smmu_tlb_inv_asid(struct arm_smmu_device *smmu, u16 asid);
> -void arm_smmu_tlb_inv_range_asid(unsigned long iova, size_t size, int asid,
> -				 size_t granule, bool leaf,
> -				 struct arm_smmu_domain *smmu_domain);
> +void arm_smmu_tlb_inv_range_s1(unsigned long iova, size_t size, size_t granule,
> +			       bool leaf, struct arm_smmu_domain *smmu_domain);
> +static inline void arm_smmu_tlb_inv_all_s1(struct arm_smmu_domain *smmu_domain)
> +{
> +	arm_smmu_tlb_inv_range_s1(0, 0, PAGE_SIZE, false, smmu_domain);
> +}
> +
>   int arm_smmu_atc_inv_domain(struct arm_smmu_domain *smmu_domain,
>   			    unsigned long iova, size_t size);
>
  
Jason Gunthorpe Jan. 18, 2024, 2:22 p.m. UTC | #3
On Thu, Jan 18, 2024 at 07:21:10PM +0800, zhangzekun (A) wrote:
> > What is the rest of the call chain? How did you get into such a large
> > invalidation?
> we are doing some test with qemu virtual machines. The testing platform has
> 96 cores and 960GB memories. It will enable 64 virtual machines and every
> virtual machine will cost 8G memories. When batching closing these 64
> virtaul machine concurrently, we will get soft lockup.
> 
>  WARN: soft lockup - CPU#18 stuck for 11s! [worker:3219]
> ....
> Call trace:
>  dump_backtrace+0x0/0x200
>  show_stack+0x20/0x30
>  dump_stack+0xf0/0x138
>  watchdog_print_info+0x48/0x54
>  watchdog_process_before_softlockup+0x9c/0xa0
>  watchdog_timer_fn+0x1ac/0x2f0
>  __run_hrtimer+0x98/0x2b4
>  __hrtimer_run_queues+0xc0/0x13c
>  hrtimer_interrupt+0x150/0x3e4
>  arch_timer_handler_phys+0x3c/0x50
>  handle_percpu_devid_irq+0x90/0x1f4
>  __handle_domain_irq+0x84/0xfc
>  gic_handle_irq+0x88/0x2b0
>  el1_irq+0xb8/0x140
>  ktime_get+0x3c/0xac
>  arm_smmu_cmdq_poll_until_not_full+0xb0/0x1a0
>  arm_smmu_cmdq_issue_cmdlist+0x194/0x5f4
>  __arm_smmu_tlb_inv_range+0x114/0x22c
>  arm_smmu_tlb_inv_walk+0x88/0x120
>  __arm_lpae_unmap+0x188/0x2c0
>  __arm_lpae_unmap+0x104/0x2c0
>  arm_lpae_unmap+0x68/0x80
>  arm_smmu_unmap+0x24/0x40
>  __iommu_unmap+0xd8/0x210
>  iommu_unmap_fast+0x18/0x24
>  unmap_unpin_fast+0x7c/0x140 [vfio_iommu_type1]
>  vfio_unmap_unpin+0x14c/0x3ac [vfio_iommu_type1]
>  vfio_remove_dma+0x38/0x124 [vfio_iommu_type1]
>  vfio_iommu_type1_detach_group+0x4b8/0x4e0 [vfio_iommu_type1]
>  __vfio_group_unset_container+0x58/0x18c [vfio]
>  vfio_group_try_dissolve_container+0x80/0x94 [vfio]
>  vfio_group_put_external_user+0x20/0x54 [vfio]
>  kvm_vfio_destroy+0xa8/0x12c
>  kvm_destroy_vm+0x20c/0x300
>  kvm_put_kvm+0x74/0xa0
>  kvm_vcpu_release+0x20/0x30
>  __fput+0xc4/0x270
>  ____fput+0x18/0x24
>  task_work_run+0xd0/0x1a0
>  do_exit+0x1d8/0x480
>  do_group_exit+0x40/0x130
>  get_signal+0x1f8/0x744
>  do_signal+0x98/0x204
>  do_notify_resume+0x15c/0x1e0
>  work_pending+0xc/0xa4

I see, that is interesting. 

I think iommufd is close to substantialy improving this. It already
has the domain destruction ordering:
 - Remove all domain attachments
 - Read back phys addrs, unmap and unpin
 - iommu_domain_free()

Currently smmu continues to issue ASID invalidations for the domain
even though there no references to it. However this is pretty
pointless here as we are going to free the ASID right away.

I'm going to make a series to allow smmu to support multi-instances on
a single domain. In this case when the domain looses all its
attachments it looses all its instances too and it won't have to do
*anything* for invalidation in this workflow.

Ie your flow will boil down to a single ASID invalidation once the
domain attachments are all removed, then on invalidations during unmap
and no invaludation at free. This would be a significantly faster
teardown I suspect.

> Thanks for your patch, the patch described below make sense to me. After
> spliting TLB invalidate operations according to domain stage,
> arm_smmu_tlb_inv_*_s1() can be used both in SVA codes and original smmu
> codes, and can implement arm_smmu_tlb_inv_all_s1() just using
> arm_smmu_tlb_inv_range_s1() by passing a special size. But the current
> smmu-v3 driver has not move "asid" into "struct smmu_domain", we still need
> to pass the exact asid from SVA.

Yes, it is not intended to be applied on the current kernel, there are
a lot of structural problems with how SVA works there..

Jason
  
Robin Murphy Jan. 24, 2024, 12:17 p.m. UTC | #4
On 18/01/2024 2:22 pm, Jason Gunthorpe wrote:
> On Thu, Jan 18, 2024 at 07:21:10PM +0800, zhangzekun (A) wrote:
>>> What is the rest of the call chain? How did you get into such a large
>>> invalidation?
>> we are doing some test with qemu virtual machines. The testing platform has
>> 96 cores and 960GB memories. It will enable 64 virtual machines and every
>> virtual machine will cost 8G memories. When batching closing these 64
>> virtaul machine concurrently, we will get soft lockup.
>>
>>   WARN: soft lockup - CPU#18 stuck for 11s! [worker:3219]
>> ....
>> Call trace:
>>   dump_backtrace+0x0/0x200
>>   show_stack+0x20/0x30
>>   dump_stack+0xf0/0x138
>>   watchdog_print_info+0x48/0x54
>>   watchdog_process_before_softlockup+0x9c/0xa0
>>   watchdog_timer_fn+0x1ac/0x2f0
>>   __run_hrtimer+0x98/0x2b4
>>   __hrtimer_run_queues+0xc0/0x13c
>>   hrtimer_interrupt+0x150/0x3e4
>>   arch_timer_handler_phys+0x3c/0x50
>>   handle_percpu_devid_irq+0x90/0x1f4
>>   __handle_domain_irq+0x84/0xfc
>>   gic_handle_irq+0x88/0x2b0
>>   el1_irq+0xb8/0x140
>>   ktime_get+0x3c/0xac
>>   arm_smmu_cmdq_poll_until_not_full+0xb0/0x1a0
>>   arm_smmu_cmdq_issue_cmdlist+0x194/0x5f4
>>   __arm_smmu_tlb_inv_range+0x114/0x22c
>>   arm_smmu_tlb_inv_walk+0x88/0x120
>>   __arm_lpae_unmap+0x188/0x2c0
>>   __arm_lpae_unmap+0x104/0x2c0
>>   arm_lpae_unmap+0x68/0x80
>>   arm_smmu_unmap+0x24/0x40
>>   __iommu_unmap+0xd8/0x210
>>   iommu_unmap_fast+0x18/0x24
>>   unmap_unpin_fast+0x7c/0x140 [vfio_iommu_type1]
>>   vfio_unmap_unpin+0x14c/0x3ac [vfio_iommu_type1]
>>   vfio_remove_dma+0x38/0x124 [vfio_iommu_type1]
>>   vfio_iommu_type1_detach_group+0x4b8/0x4e0 [vfio_iommu_type1]
>>   __vfio_group_unset_container+0x58/0x18c [vfio]
>>   vfio_group_try_dissolve_container+0x80/0x94 [vfio]
>>   vfio_group_put_external_user+0x20/0x54 [vfio]
>>   kvm_vfio_destroy+0xa8/0x12c
>>   kvm_destroy_vm+0x20c/0x300
>>   kvm_put_kvm+0x74/0xa0
>>   kvm_vcpu_release+0x20/0x30
>>   __fput+0xc4/0x270
>>   ____fput+0x18/0x24
>>   task_work_run+0xd0/0x1a0
>>   do_exit+0x1d8/0x480
>>   do_group_exit+0x40/0x130
>>   get_signal+0x1f8/0x744
>>   do_signal+0x98/0x204
>>   do_notify_resume+0x15c/0x1e0
>>   work_pending+0xc/0xa4
> 
> I see, that is interesting.
> 
> I think iommufd is close to substantialy improving this. It already
> has the domain destruction ordering:
>   - Remove all domain attachments
>   - Read back phys addrs, unmap and unpin
>   - iommu_domain_free()

I don't think the story here has changed from when I first remember this 
being discussed probably 7 or 8 years ago - what VFIO has always really 
wanted for this situation is to know when it's doing a complete teardown 
and just detach and free the domain, then unpin the pages afterwards, 
without wasting any time at all on frivolous unmapping. However so far 
it's worked out that minor low-level tweaks like the unmap_fast API have 
kept kept the overhead tolerable enough that nobody's been inspired to 
attempt the bigger task.

However for SMMU in particular it's almost certainly confounded by 
io_pgtable_tlb_flush_walk() being a bit rubbish for the situation that I 
bet is happening here - if we don't have range invalidate, then 
pessimistically invalidating a 1GB table at 4KB granularity without even 
looking isn't likely to save much time over doing a more accurate 
recursive invalidation even in the worst case that the whole thing 
*does* turn out to be mapped as pages. But in the likely case that there 
are at least some intermediate-level block mappings in there, it's an 
instant win. I recall I had along-standing to-do item to get rid of 
tlb_flush_walk from io-pgtable altogether, although that may have been 
tied in to the freelist stuff; I'd need to find time to page it all back 
in to remember exactly why...

Thanks,
Robin.

> Currently smmu continues to issue ASID invalidations for the domain
> even though there no references to it. However this is pretty
> pointless here as we are going to free the ASID right away.
> 
> I'm going to make a series to allow smmu to support multi-instances on
> a single domain. In this case when the domain looses all its
> attachments it looses all its instances too and it won't have to do
> *anything* for invalidation in this workflow.
> 
> Ie your flow will boil down to a single ASID invalidation once the
> domain attachments are all removed, then on invalidations during unmap
> and no invaludation at free. This would be a significantly faster
> teardown I suspect.
> 
>> Thanks for your patch, the patch described below make sense to me. After
>> spliting TLB invalidate operations according to domain stage,
>> arm_smmu_tlb_inv_*_s1() can be used both in SVA codes and original smmu
>> codes, and can implement arm_smmu_tlb_inv_all_s1() just using
>> arm_smmu_tlb_inv_range_s1() by passing a special size. But the current
>> smmu-v3 driver has not move "asid" into "struct smmu_domain", we still need
>> to pass the exact asid from SVA.
> 
> Yes, it is not intended to be applied on the current kernel, there are
> a lot of structural problems with how SVA works there..
> 
> Jason
  
Jason Gunthorpe Jan. 24, 2024, 4:28 p.m. UTC | #5
On Wed, Jan 24, 2024 at 12:17:43PM +0000, Robin Murphy wrote:

> I don't think the story here has changed from when I first remember this
> being discussed probably 7 or 8 years ago - what VFIO has always really
> wanted for this situation is to know when it's doing a complete teardown and
> just detach and free the domain, then unpin the pages afterwards, without
> wasting any time at all on frivolous unmapping. 

What VFIO/iommufd really wants is a fast way to to read back and unmap
the PFNs from the domain. Today it has to call iommu_iova_to_phys()
and then unmap which ends up walking the radix many times.

Ideally we'd batch read and unmap in a single call with a single radix
walk.

This is because the domain is used to store the 'struct page *' which
is needed to unpin the memory.

Combined with the no-attachments optimization I described it could be
pretty close to optimal.

Jason
  

Patch

diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
index 05722121f00e..164a218a4d41 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c
@@ -203,15 +203,6 @@  static void arm_smmu_free_shared_cd(struct arm_smmu_ctx_desc *cd)
 	}
 }
 
-/*
- * Cloned from the MAX_TLBI_OPS in arch/arm64/include/asm/tlbflush.h, this
- * is used as a threshold to replace per-page TLBI commands to issue in the
- * command queue with an address-space TLBI command, when SMMU w/o a range
- * invalidation feature handles too many per-page TLBI commands, which will
- * otherwise result in a soft lockup.
- */
-#define CMDQ_MAX_TLBI_OPS		(1 << (PAGE_SHIFT - 3))
-
 static void arm_smmu_mm_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
 						struct mm_struct *mm,
 						unsigned long start,
@@ -228,7 +219,7 @@  static void arm_smmu_mm_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
 	 */
 	size = end - start;
 	if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_RANGE_INV)) {
-		if (size >= CMDQ_MAX_TLBI_OPS * PAGE_SIZE)
+		if (size >= CMDQ_MAX_TLBI_OPS(PAGE_SIZE) * PAGE_SIZE)
 			size = 0;
 	} else {
 		if (size == ULONG_MAX)
diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
index 0ffb1cf17e0b..cecccba17511 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
@@ -1997,6 +1997,14 @@  static void arm_smmu_tlb_inv_page_nosync(struct iommu_iotlb_gather *gather,
 static void arm_smmu_tlb_inv_walk(unsigned long iova, size_t size,
 				  size_t granule, void *cookie)
 {
+	struct arm_smmu_domain *smmu_domain = cookie;
+	struct arm_smmu_device *smmu = smmu_domain->smmu;
+
+	if (!(smmu->features & ARM_SMMU_FEAT_RANGE_INV) &&
+	    size >= CMDQ_MAX_TLBI_OPS(granule) * granule) {
+		arm_smmu_tlb_inv_context(cookie);
+		return;
+	}
 	arm_smmu_tlb_inv_range_domain(iova, size, granule, false, cookie);
 }
 
@@ -2502,13 +2510,20 @@  static void arm_smmu_iotlb_sync(struct iommu_domain *domain,
 				struct iommu_iotlb_gather *gather)
 {
 	struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain);
+	struct arm_smmu_device *smmu = smmu_domain->smmu;
+	size_t size = gather->end - gather->start + 1;
+	size_t granule = gather->pgsize;
 
-	if (!gather->pgsize)
+	if (!granule)
 		return;
 
-	arm_smmu_tlb_inv_range_domain(gather->start,
-				      gather->end - gather->start + 1,
-				      gather->pgsize, true, smmu_domain);
+	if (!(smmu->features & ARM_SMMU_FEAT_RANGE_INV) &&
+	    size >= CMDQ_MAX_TLBI_OPS(granule) * granule) {
+		arm_smmu_tlb_inv_context(smmu_domain);
+		return;
+	}
+	arm_smmu_tlb_inv_range_domain(gather->start, size,
+				      granule, true, smmu_domain);
 }
 
 static phys_addr_t
diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
index 65fb388d5173..a9a7376c0437 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
@@ -431,6 +431,16 @@  struct arm_smmu_ste {
 #define MSI_IOVA_BASE			0x8000000
 #define MSI_IOVA_LENGTH			0x100000
 
+/*
+ * Similar to MAX_DVM_OPS in arch/arm64/include/asm/tlbflush.h, this is used
+ * as a threshold to replace per-page TLBI commands to issue in the command
+ * queue with an address-space TLBI command, when SMMU w/o a range invalidation
+ * feature handles too many per-page TLBI commands, which will otherwise result
+ * in a soft lockup.
+ */
+
+#define CMDQ_MAX_TLBI_OPS(granule)    (1 << (ilog2(granule) - 3))
+
 enum pri_resp {
 	PRI_RESP_DENY = 0,
 	PRI_RESP_FAIL = 1,