All of lore.kernel.org
 help / color / mirror / Atom feed
From: Jason Gunthorpe <jgg@nvidia.com>
To: Michael Shavit <mshavit@google.com>
Cc: iommu@lists.linux.dev, Joerg Roedel <joro@8bytes.org>,
	linux-arm-kernel@lists.infradead.org,
	Robin Murphy <robin.murphy@arm.com>,
	Will Deacon <will@kernel.org>,
	Jean-Philippe Brucker <jean-philippe@linaro.org>,
	Nicolin Chen <nicolinc@nvidia.com>
Subject: Re: [PATCH 14/27] iommu/arm-smmu-v3: Make changing domains be hitless for ATS
Date: Thu, 26 Oct 2023 11:38:09 -0300	[thread overview]
Message-ID: <20231026143809.GQ3952@nvidia.com> (raw)
In-Reply-To: <CAKHBV27tKWzaoVy5BaVJG1VHPR=duU7d08c6J=RvYsO8hiLNcQ@mail.gmail.com>

On Thu, Oct 26, 2023 at 03:00:11PM +0800, Michael Shavit wrote:

> Huh. Can this be checked by calling iommu_get_domain_for_dev() and
> comparing against the input iommu_domain parameter, or is
> iommu_get_domain_for_dev() cleared in this scenario?

Yes, that should work, but I dislike the frailty. 

Hmm. When I first wrote this the whole picture wasn't finished, now I
think we can just take the optimization out. If the same master has
two master_domains in the list for a moment it is not actually a
problem. The commit stage will remove only one.

So, like this:

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 15d1688d7f7034..31ce450448a7f0 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
@@ -2490,6 +2490,10 @@ static void arm_smmu_remove_master_domain(struct arm_smmu_master *master,
 	struct arm_smmu_master_domain *master_domain;
 	unsigned long flags;
 
+	/* NULL means the old domain is IDENTITY/BLOCKED which we don't track */
+	if (!smmu_domain)
+                return;
+
 	spin_lock_irqsave(&smmu_domain->devices_lock, flags);
 	master_domain = arm_smmu_find_master_domain(smmu_domain, master);
 	if (master_domain) {
@@ -2502,8 +2506,7 @@ static void arm_smmu_remove_master_domain(struct arm_smmu_master *master,
 }
 
 struct attach_state {
-	bool existing_master_domain : 1;
-	bool want_ats : 1;
+	bool want_ats;
 };
 
 /*
@@ -2514,7 +2517,6 @@ static int arm_smmu_attach_prepare(struct arm_smmu_master *master,
 				   struct arm_smmu_domain *smmu_domain,
 				   struct attach_state *state)
 {
-	struct arm_smmu_master_domain *cur_master_domain;
 	struct arm_smmu_master_domain *master_domain;
 	unsigned long flags;
 
@@ -2541,17 +2543,14 @@ static int arm_smmu_attach_prepare(struct arm_smmu_master *master,
 	 * It is tempting to make this list only track masters that are using
 	 * ATS, but arm_smmu_share_asid() also uses this to change the ASID of a
 	 * domain, unrelated to ATS.
+	 *
+	 * Notice if we are re-attaching the same domain then the list will have
+	 * two identical entries and commit will remove only one of them.
 	 */
 	spin_lock_irqsave(&smmu_domain->devices_lock, flags);
-	cur_master_domain = arm_smmu_find_master_domain(smmu_domain, master);
-	if (cur_master_domain) {
-		kfree(master_domain);
-		state->existing_master_domain = true;
-	} else {
-		if (state->want_ats)
-			atomic_inc(&smmu_domain->nr_ats_masters);
-		list_add(&master_domain->devices_elm, &smmu_domain->devices);
-	}
+	if (state->want_ats)
+		atomic_inc(&smmu_domain->nr_ats_masters);
+	list_add(&master_domain->devices_elm, &smmu_domain->devices);
 	spin_unlock_irqrestore(&smmu_domain->devices_lock, flags);
 	return 0;
 }
@@ -2581,13 +2580,9 @@ static void arm_smmu_attach_commit(struct arm_smmu_master *master,
 		arm_smmu_atc_inv_master(master);
 	}
 
-	if (!state->existing_master_domain) {
-		struct arm_smmu_domain *old_smmu_domain = to_smmu_domain_safe(
-			iommu_get_domain_for_dev(master->dev));
-
-		if (old_smmu_domain)
-			arm_smmu_remove_master_domain(master, old_smmu_domain);
-	}
+	arm_smmu_remove_master_domain(
+		master,
+		to_smmu_domain_safe(iommu_get_domain_for_dev(master->dev)));
 }
 
 static int arm_smmu_attach_dev(struct iommu_domain *domain, struct device *dev)

WARNING: multiple messages have this Message-ID (diff)
From: Jason Gunthorpe <jgg@nvidia.com>
To: Michael Shavit <mshavit@google.com>
Cc: iommu@lists.linux.dev, Joerg Roedel <joro@8bytes.org>,
	linux-arm-kernel@lists.infradead.org,
	Robin Murphy <robin.murphy@arm.com>,
	Will Deacon <will@kernel.org>,
	Jean-Philippe Brucker <jean-philippe@linaro.org>,
	Nicolin Chen <nicolinc@nvidia.com>
Subject: Re: [PATCH 14/27] iommu/arm-smmu-v3: Make changing domains be hitless for ATS
Date: Thu, 26 Oct 2023 11:38:09 -0300	[thread overview]
Message-ID: <20231026143809.GQ3952@nvidia.com> (raw)
In-Reply-To: <CAKHBV27tKWzaoVy5BaVJG1VHPR=duU7d08c6J=RvYsO8hiLNcQ@mail.gmail.com>

On Thu, Oct 26, 2023 at 03:00:11PM +0800, Michael Shavit wrote:

> Huh. Can this be checked by calling iommu_get_domain_for_dev() and
> comparing against the input iommu_domain parameter, or is
> iommu_get_domain_for_dev() cleared in this scenario?

Yes, that should work, but I dislike the frailty. 

Hmm. When I first wrote this the whole picture wasn't finished, now I
think we can just take the optimization out. If the same master has
two master_domains in the list for a moment it is not actually a
problem. The commit stage will remove only one.

So, like this:

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 15d1688d7f7034..31ce450448a7f0 100644
--- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
+++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
@@ -2490,6 +2490,10 @@ static void arm_smmu_remove_master_domain(struct arm_smmu_master *master,
 	struct arm_smmu_master_domain *master_domain;
 	unsigned long flags;
 
+	/* NULL means the old domain is IDENTITY/BLOCKED which we don't track */
+	if (!smmu_domain)
+                return;
+
 	spin_lock_irqsave(&smmu_domain->devices_lock, flags);
 	master_domain = arm_smmu_find_master_domain(smmu_domain, master);
 	if (master_domain) {
@@ -2502,8 +2506,7 @@ static void arm_smmu_remove_master_domain(struct arm_smmu_master *master,
 }
 
 struct attach_state {
-	bool existing_master_domain : 1;
-	bool want_ats : 1;
+	bool want_ats;
 };
 
 /*
@@ -2514,7 +2517,6 @@ static int arm_smmu_attach_prepare(struct arm_smmu_master *master,
 				   struct arm_smmu_domain *smmu_domain,
 				   struct attach_state *state)
 {
-	struct arm_smmu_master_domain *cur_master_domain;
 	struct arm_smmu_master_domain *master_domain;
 	unsigned long flags;
 
@@ -2541,17 +2543,14 @@ static int arm_smmu_attach_prepare(struct arm_smmu_master *master,
 	 * It is tempting to make this list only track masters that are using
 	 * ATS, but arm_smmu_share_asid() also uses this to change the ASID of a
 	 * domain, unrelated to ATS.
+	 *
+	 * Notice if we are re-attaching the same domain then the list will have
+	 * two identical entries and commit will remove only one of them.
 	 */
 	spin_lock_irqsave(&smmu_domain->devices_lock, flags);
-	cur_master_domain = arm_smmu_find_master_domain(smmu_domain, master);
-	if (cur_master_domain) {
-		kfree(master_domain);
-		state->existing_master_domain = true;
-	} else {
-		if (state->want_ats)
-			atomic_inc(&smmu_domain->nr_ats_masters);
-		list_add(&master_domain->devices_elm, &smmu_domain->devices);
-	}
+	if (state->want_ats)
+		atomic_inc(&smmu_domain->nr_ats_masters);
+	list_add(&master_domain->devices_elm, &smmu_domain->devices);
 	spin_unlock_irqrestore(&smmu_domain->devices_lock, flags);
 	return 0;
 }
@@ -2581,13 +2580,9 @@ static void arm_smmu_attach_commit(struct arm_smmu_master *master,
 		arm_smmu_atc_inv_master(master);
 	}
 
-	if (!state->existing_master_domain) {
-		struct arm_smmu_domain *old_smmu_domain = to_smmu_domain_safe(
-			iommu_get_domain_for_dev(master->dev));
-
-		if (old_smmu_domain)
-			arm_smmu_remove_master_domain(master, old_smmu_domain);
-	}
+	arm_smmu_remove_master_domain(
+		master,
+		to_smmu_domain_safe(iommu_get_domain_for_dev(master->dev)));
 }
 
 static int arm_smmu_attach_dev(struct iommu_domain *domain, struct device *dev)

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

  reply	other threads:[~2023-10-26 14:38 UTC|newest]

Thread overview: 88+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-10-11 23:25 [PATCH 00/27] Update SMMUv3 to the modern iommu API (part 2/2) Jason Gunthorpe
2023-10-11 23:25 ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 01/27] iommu/arm-smmu-v3: Check that the RID domain is S1 in SVA Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 02/27] iommu/arm-smmu-v3: Do not allow a SVA domain to be set on the wrong PASID Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 03/27] iommu/arm-smmu-v3: Do not ATC invalidate the entire domain Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 04/27] iommu/arm-smmu-v3: Add a type for the CD entry Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 05/27] iommu/arm-smmu-v3: Make CD programming use arm_smmu_write_entry_step() Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 06/27] iommu/arm-smmu-v3: Consolidate clearing a CD table entry Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 07/27] iommu/arm-smmu-v3: Move the CD generation for S1 domains into a function Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 08/27] iommu/arm-smmu-v3: Move allocation of the cdtable into arm_smmu_get_cd_ptr() Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 09/27] iommu/arm-smmu-v3: Allocate the CD table entry in advance Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 10/27] iommu/arm-smmu-v3: Move the CD generation for SVA into a function Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-24  4:12   ` Michael Shavit
2023-10-24  4:12     ` Michael Shavit
2023-10-24 11:52     ` Jason Gunthorpe
2023-10-24 11:52       ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 11/27] iommu/arm-smmu-v3: Lift CD programming out of the SVA notifier code Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-24  6:34   ` Michael Shavit
2023-10-24  6:34     ` Michael Shavit
2023-10-24 23:46     ` Jason Gunthorpe
2023-10-24 23:46       ` Jason Gunthorpe
2023-10-26  7:31       ` Michael Shavit
2023-10-26  7:31         ` Michael Shavit
2023-10-26 14:11         ` Jason Gunthorpe
2023-10-26 14:11           ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 12/27] iommu/arm-smmu-v3: Build the whole CD in arm_smmu_make_s1_cd() Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 13/27] iommu/arm-smmu-v3: Make smmu_domain->devices into an allocated list Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 14/27] iommu/arm-smmu-v3: Make changing domains be hitless for ATS Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-24  8:09   ` Michael Shavit
2023-10-24  8:09     ` Michael Shavit
2023-10-24 23:56     ` Jason Gunthorpe
2023-10-24 23:56       ` Jason Gunthorpe
2023-10-26  7:00       ` Michael Shavit
2023-10-26  7:00         ` Michael Shavit
2023-10-26 14:38         ` Jason Gunthorpe [this message]
2023-10-26 14:38           ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 15/27] iommu/arm-smmu-v3: Add ssid to struct arm_smmu_master_domain Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 16/27] iommu/arm-smmu-v3: Keep track of valid CD entries in the cd_table Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 17/27] iommu/arm-smmu-v3: Thread SSID through the arm_smmu_attach_*() interface Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-25 14:01   ` Jason Gunthorpe
2023-10-25 14:01     ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 18/27] iommu/arm-smmu-v3: Make SVA allocate a normal arm_smmu_domain Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-24  8:58   ` Michael Shavit
2023-10-24  8:58     ` Michael Shavit
2023-10-24 13:05     ` Jason Gunthorpe
2023-10-24 13:05       ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 19/27] iommu/arm-smmu-v3: Keep track of arm_smmu_master_domain for SVA Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 20/27] iommu: Add ops->domain_alloc_sva() Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 21/27] iommu/arm-smmu-v3: Put the SVA mmu notifier in the smmu_domain Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-25 13:56   ` Jason Gunthorpe
2023-10-25 13:56     ` Jason Gunthorpe
2023-10-25 16:23   ` Jason Gunthorpe
2023-10-25 16:23     ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 22/27] iommu/arm-smmu-v3: Consolidate freeing the ASID/VMID Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:25 ` [PATCH 23/27] iommu/arm-smmu-v3: Move the arm_smmu_asid_xa to per-smmu like vmid Jason Gunthorpe
2023-10-11 23:25   ` Jason Gunthorpe
2023-10-11 23:26 ` [PATCH 24/27] iommu/arm-smmu-v3: Bring back SVA BTM support Jason Gunthorpe
2023-10-11 23:26   ` Jason Gunthorpe
2023-10-11 23:26 ` [PATCH 25/27] iommu/arm-smmu-v3: Allow IDENTITY/BLOCKED to be set while PASID is used Jason Gunthorpe
2023-10-11 23:26   ` Jason Gunthorpe
2023-10-25 15:10   ` Jason Gunthorpe
2023-10-25 15:10     ` Jason Gunthorpe
2023-10-11 23:26 ` [PATCH 26/27] iommu/arm-smmu-v3: Allow a PASID to be set when RID is IDENTITY/BLOCKED Jason Gunthorpe
2023-10-11 23:26   ` Jason Gunthorpe
2023-10-11 23:26 ` [PATCH 27/27] iommu/arm-smmu-v3: Allow setting a S1 domain to a PASID Jason Gunthorpe
2023-10-11 23:26   ` Jason Gunthorpe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20231026143809.GQ3952@nvidia.com \
    --to=jgg@nvidia.com \
    --cc=iommu@lists.linux.dev \
    --cc=jean-philippe@linaro.org \
    --cc=joro@8bytes.org \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=mshavit@google.com \
    --cc=nicolinc@nvidia.com \
    --cc=robin.murphy@arm.com \
    --cc=will@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.