[PATCH v3 0/3] dmaengine: fix kref underflow and UAF in dma_chan_put()

[PATCH v3 0/3] dmaengine: fix kref underflow and UAF in dma_chan_put()

[Shivank Garg]

Fix bugs related to dma_chan_put(), found while testing with SDXI[1].

[1]: https://lore.kernel.org/dmaengine/20260605-sdxi-base-v3-0-4d38ca2bdffe@amd.com

Signed-off-by: Shivank Garg <shivankg@amd.com>
---
Changes in v3:
- Add patch 3: add synchronize_rcu() to wait for RCU readers to prevent
  use-after-free. (Sashiko)
- Link to v2: https://lore.kernel.org/r/20260526-dmaengine-kref-fix-v2-0-3df60afac01d@amd.com

Changes in v2:
- Add patch 2 fixing the dma_chan_put()/dma_release_channel() use-after-free (sashiko) 
- Link to v1: https://lore.kernel.org/r/20260518-dmaengine-kref-fix-v1-1-4d6125048fb7@amd.com

---
Shivank Garg (3):
      dmaengine: Fix device kref underflow in dma_chan_put()
      dmaengine: fix use-after-free in dma_chan_put() and dma_release_channel()
      dmaengine: wait for RCU readers before releasing dma_device

 drivers/dma/dmaengine.c | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)
---
base-commit: 0d995da5fb97e8c312834575604d4423eb6225b7
change-id: 20260518-dmaengine-kref-fix-7b21acb09455

Best regards,
-- 
Shivank Garg <shivankg@amd.com>

[PATCH v3 1/3] dmaengine: Fix device kref underflow in dma_chan_put()

[Shivank Garg]

dma_chan_get() takes chan->device->ref only on the slow path:

	/* no kref on fast path */
	if (chan->client_count) {
		__module_get(owner);
		chan->client_count++;
		return 0;
	}
	if (!try_module_get(owner))
		return -ENODEV;
	ret = kref_get_unless_zero(&chan->device->ref);

dma_chan_put() drops the ref unconditionally, so every fast-path
get/put pair drops one extra device reference.

The bug fires when two conditions hold together: a non-private
provider has a persistent client holding chan->client_count > 0
and another client cycles dmaengine_get()/dmaengine_put().
When the kref hits zero, the subsequent dma_find_channel() returns
NULL even though the provider module is still loaded.

Fix this by dropping device->ref only on the last put, matching the
single slow-path get.

Fixes: 8ad342a86359 ("dmaengine: Add reference counting to dma_device struct")
Signed-off-by: Shivank Garg <shivankg@amd.com>
---
 drivers/dma/dmaengine.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index 6ffd8bd82154..516d6d933208 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -515,7 +515,9 @@ static void dma_chan_put(struct dma_chan *chan)
 		chan->route_data = NULL;
 	}
 
-	dma_device_put(chan->device);
+	/* This channel is not in use anymore, drop the device ref */
+	if (!chan->client_count)
+		dma_device_put(chan->device);
 	module_put(dma_chan_to_owner(chan));
 }
 

-- 
2.43.0

Re: [PATCH v3 1/3] dmaengine: Fix device kref underflow in dma_chan_put()

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:

Pre-existing issues:
- [High] dma_chan_put() accesses chan->device->owner after potentially freeing chan and chan->device via dma_device_put().
- [High] dma_release_channel() accesses `chan` fields after calling `dma_chan_put()`, which may have freed the channel.
- [High] RCU read-side critical section in `dma_issue_pending_all()` can access a synchronously freed `struct dma_device`.
--

--- Patch [1]: [PATCH v3 1/3] dmaengine: Fix device kref underflow in dma_chan_put() ---
commit 0f010c1f7452238748341b884bda3d3bba2c74ac
Author: Shivank Garg <shivankg@amd.com>

dmaengine: Fix device kref underflow in dma_chan_put()

This commit fixes a device reference underflow bug by ensuring the device 
reference is only dropped on the last put of a channel, matching the behavior 
of the get path.

> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> index 6ffd8bd82154a..516d6d9332083 100644
> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -515,7 +515,9 @@ static void dma_chan_put(struct dma_chan *chan)
>  		chan->route_data = NULL;
>  	}
>  
> -	dma_device_put(chan->device);
> +	/* This channel is not in use anymore, drop the device ref */
> +	if (!chan->client_count)
> +		dma_device_put(chan->device);
>  	module_put(dma_chan_to_owner(chan));

[Severity: High]
This is a pre-existing issue, but does accessing chan after dma_device_put()
risk a use-after-free?

If dma_device_put() drops the last reference to the DMA device, it can
trigger synchronous release of the device memory, which might also free the
embedded dma_chan.

Could dma_chan_to_owner(chan) then dereference freed memory by evaluating
chan->device->owner?

>  }

[Severity: High]
This isn't a bug introduced by this patch, but does dma_release_channel()
also access freed memory under the same conditions?

drivers/dma/dmaengine.c:dma_release_channel() {
	...
	dma_chan_put(chan);

	if (chan->slave) {
		sysfs_remove_link(&chan->dev->device.kobj, DMA_SLAVE_NAME);
	...
}

If dma_chan_put() frees the channel when the last device reference is dropped,
is it safe to access chan->slave and chan->dev afterwards to tear down
the sysfs links?

[Severity: High]
This is also a pre-existing issue, but does the RCU read side in
dma_issue_pending_all() race with synchronous device release?

dma_issue_pending_all() relies on an RCU read lock to traverse the device list:

drivers/dma/dmaengine.c:dma_issue_pending_all() {
	...
	rcu_read_lock();
	list_for_each_entry_rcu(device, &dma_device_list, global_node) {
		if (dma_has_cap(DMA_PRIVATE, device->cap_mask))
	...
}

However, dma_device_release() unlinks the device and synchronously frees
it without observing an RCU grace period:

drivers/dma/dmaengine.c:dma_device_release() {
	...
	list_del_rcu(&device->global_node);
	dma_channel_rebalance();

	if (device->device_release)
		device->device_release(device);
	...
}

Could the RCU reader dereference a struct dma_device that was just
freed by dma_device_release()?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260816-dmaengine-kref-fix-v3-0-7e76187145df@amd.com?part=1

Re: [PATCH v3 1/3] dmaengine: Fix device kref underflow in dma_chan_put()

[Garg, Shivank]

On Sun, 2026-08-16 at 16:02 +0000, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
> 
> Pre-existing issues:
> - [High] dma_chan_put() accesses chan->device->owner after potentially freeing chan and chan->device via dma_device_put().
> - [High] dma_release_channel() accesses `chan` fields after calling `dma_chan_put()`, which may have freed the channel.
> - [High] RCU read-side critical section in `dma_issue_pending_all()` can access a synchronously freed `struct dma_device`.
> --
> 
> --- Patch [1]: [PATCH v3 1/3] dmaengine: Fix device kref underflow in dma_chan_put() ---
> commit 0f010c1f7452238748341b884bda3d3bba2c74ac
> Author: Shivank Garg <shivankg@amd.com>
> 
> dmaengine: Fix device kref underflow in dma_chan_put()
> 
> This commit fixes a device reference underflow bug by ensuring the device 
> reference is only dropped on the last put of a channel, matching the behavior 
> of the get path.
> 
> > diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> > index 6ffd8bd82154a..516d6d9332083 100644
> > --- a/drivers/dma/dmaengine.c
> > +++ b/drivers/dma/dmaengine.c
> > @@ -515,7 +515,9 @@ static void dma_chan_put(struct dma_chan *chan)
> >  		chan->route_data = NULL;
> >  	}
> >  
> > -	dma_device_put(chan->device);
> > +	/* This channel is not in use anymore, drop the device ref */
> > +	if (!chan->client_count)
> > +		dma_device_put(chan->device);
> >  	module_put(dma_chan_to_owner(chan));
> 
> [Severity: High]
> This is a pre-existing issue, but does accessing chan after dma_device_put()
> risk a use-after-free?
> 
> If dma_device_put() drops the last reference to the DMA device, it can
> trigger synchronous release of the device memory, which might also free the
> embedded dma_chan.
> 
> Could dma_chan_to_owner(chan) then dereference freed memory by evaluating
> chan->device->owner?
> 

I'm already fixing this in patch 2 by caching chan->device->owner.

> >  }
> 
> [Severity: High]
> This isn't a bug introduced by this patch, but does dma_release_channel()
> also access freed memory under the same conditions?
> 
> drivers/dma/dmaengine.c:dma_release_channel() {
> 	...
> 	dma_chan_put(chan);
> 
> 	if (chan->slave) {
> 		sysfs_remove_link(&chan->dev->device.kobj, DMA_SLAVE_NAME);
> 	...
> }
> 
> If dma_chan_put() frees the channel when the last device reference is dropped,
> is it safe to access chan->slave and chan->dev afterwards to tear down
> the sysfs links?

I addressed this in patch 2, but Sashiko pointed the fix would cause NULL
dereference in stm32_dma3_pm_suspend().
> 
> [Severity: High]
> This is also a pre-existing issue, but does the RCU read side in
> dma_issue_pending_all() race with synchronous device release?
> 
> dma_issue_pending_all() relies on an RCU read lock to traverse the device list:
> 
> drivers/dma/dmaengine.c:dma_issue_pending_all() {
> 	...
> 	rcu_read_lock();
> 	list_for_each_entry_rcu(device, &dma_device_list, global_node) {
> 		if (dma_has_cap(DMA_PRIVATE, device->cap_mask))
> 	...
> }
> 
> However, dma_device_release() unlinks the device and synchronously frees
> it without observing an RCU grace period:
> 
> drivers/dma/dmaengine.c:dma_device_release() {
> 	...
> 	list_del_rcu(&device->global_node);
> 	dma_channel_rebalance();
> 
> 	if (device->device_release)
> 		device->device_release(device);
> 	...
> }
> 
> Could the RCU reader dereference a struct dma_device that was just
> freed by dma_device_release()?
> 
Fixed in Patch 3 by adding synchronize_rcu().

Thanks,
Shivank

[PATCH v3 2/3] dmaengine: fix use-after-free in dma_chan_put() and dma_release_channel()

[Shivank Garg]

When dma_device_put() drops the last reference on chan->device->ref,
dma_device_release() runs and may free the dma_device along with its
channels.

Two paths still read that memory after the put:
 - dma_chan_put() reads chan->device->owner via dma_chan_to_owner()
   for the trailing module_put().
 - dma_release_channel() calls dma_chan_put() before reading chan->slave,
   chan->name, chan->dev and chan->dbg_client_name.

KASAN catches the first one:

	slab-use-after-free in dma_chan_put+0x3e6/0x4c0
	Read of size 8 by task insmod/6319
	Freed by task 6319:
	  kfree+0x225/0x470
	  dma_chan_put+0x395/0x4c0
	  dmaengine_put+0xf8/0x160

Cache the module owner in dma_chan_put() before the put so the trailing
module_put() does not need chan->device. In dma_release_channel(), move
dma_chan_put() to the end, after every chan/device read.

Fixes: 8ad342a86359 ("dmaengine: Add reference counting to dma_device struct")
Suggested-by: Sashiko <sashiko-bot@kernel.org>
Link: https://sashiko.dev/#/patchset/20260518-dmaengine-kref-fix-v1-1-4d6125048fb7@amd.com
Signed-off-by: Shivank Garg <shivankg@amd.com>
---
 drivers/dma/dmaengine.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index 516d6d933208..bf491eb10596 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -495,10 +495,13 @@ static int dma_chan_get(struct dma_chan *chan)
  */
 static void dma_chan_put(struct dma_chan *chan)
 {
+	struct module *owner;
+
 	/* This channel is not in use, bail out */
 	if (!chan->client_count)
 		return;
 
+	owner = dma_chan_to_owner(chan);
 	chan->client_count--;
 
 	/* This channel is not in use anymore, free it */
@@ -518,7 +521,7 @@ static void dma_chan_put(struct dma_chan *chan)
 	/* This channel is not in use anymore, drop the device ref */
 	if (!chan->client_count)
 		dma_device_put(chan->device);
-	module_put(dma_chan_to_owner(chan));
+	module_put(owner);
 }
 
 enum dma_status dma_sync_wait(struct dma_chan *chan, dma_cookie_t cookie)
@@ -916,8 +919,6 @@ void dma_release_channel(struct dma_chan *chan)
 	if (--chan->device->privatecnt == 0)
 		dma_cap_clear(DMA_PRIVATE, chan->device->cap_mask);
 
-	dma_chan_put(chan);
-
 	if (chan->slave) {
 		sysfs_remove_link(&chan->dev->device.kobj, DMA_SLAVE_NAME);
 		sysfs_remove_link(&chan->slave->kobj, chan->name);
@@ -930,6 +931,7 @@ void dma_release_channel(struct dma_chan *chan)
 	kfree(chan->dbg_client_name);
 	chan->dbg_client_name = NULL;
 #endif
+	dma_chan_put(chan);
 	mutex_unlock(&dma_list_mutex);
 }
 EXPORT_SYMBOL_GPL(dma_release_channel);

-- 
2.43.0

Re: [PATCH v3 2/3] dmaengine: fix use-after-free in dma_chan_put() and dma_release_channel()

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Clearing `chan->slave` before stopping the hardware introduces a NULL pointer dereference in `stm32_dma3_pm_suspend()`.
--

--- Patch [2]: [PATCH v3 2/3] dmaengine: fix use-after-free in dma_chan_put() and dma_release_channel() ---
commit c0bca60dc6977cca7352bacd1095081d89c88211
Author: Shivank Garg <shivankg@amd.com>

dmaengine: fix use-after-free in dma_chan_put() and dma_release_channel()

When dma_device_put() drops the last reference on chan->device->ref,
dma_device_release() may free the dma_device. This commit caches the
module owner and delays dma_chan_put() to prevent reading freed memory.

Link: https://sashiko.dev/#/patchset/20260518-dmaengine-kref-fix-v1-1-4d6125048fb7@amd.com

> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -916,8 +919,6 @@ void dma_release_channel(struct dma_chan *chan)
>  	if (--chan->device->privatecnt == 0)
>  		dma_cap_clear(DMA_PRIVATE, chan->device->cap_mask);
>  
> -	dma_chan_put(chan);
> -
>  	if (chan->slave) {
>  		sysfs_remove_link(&chan->dev->device.kobj, DMA_SLAVE_NAME);
>  		sysfs_remove_link(&chan->slave->kobj, chan->name);
>  		kfree(chan->name);
>  		chan->name = NULL;
>  		chan->slave = NULL;
>  	}
>  
>  #ifdef CONFIG_DEBUG_FS
>  	kfree(chan->dbg_client_name);
>  	chan->dbg_client_name = NULL;
>  #endif
> +	dma_chan_put(chan);

[Severity: High]
Could moving dma_chan_put() here introduce a NULL pointer dereference
during system suspend?

By moving dma_chan_put() after the block that clears chan->slave, the
hardware channel may still be active while chan->slave is NULL. If a
system suspend occurs concurrently, drivers like stm32-dma3 might check
the hardware state and attempt to print the slave device name:

drivers/dma/stm32/stm32-dma3.c:stm32_dma3_pm_suspend() {
    ...
	ccr = readl_relaxed(ddata->base + STM32_DMA3_CCR(chan->id));
	if (ccr & CCR_EN) {
		dev_warn(dev, "Suspend is prevented: %s still in use by %s\n",
			 dma_chan_name(c), dev_name(c->slave));
    ...
}

Since dma_chan_put() (which stops the hardware) hasn't been called yet,
CCR_EN would still be active, and dev_name(c->slave) would dereference
the NULL pointer we just set. 

Is there a way to stop the channel before clearing the slave pointer
without triggering the original use-after-free?

>  	mutex_unlock(&dma_list_mutex);
>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260816-dmaengine-kref-fix-v3-0-7e76187145df@amd.com?part=2

Re: [PATCH v3 2/3] dmaengine: fix use-after-free in dma_chan_put() and dma_release_channel()

[Garg, Shivank]

On Sun, 2026-08-16 at 16:07 +0000, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [High] Clearing `chan->slave` before stopping the hardware introduces a NULL pointer dereference in `stm32_dma3_pm_suspend()`.
> --
> 
> --- Patch [2]: [PATCH v3 2/3] dmaengine: fix use-after-free in dma_chan_put() and dma_release_channel() ---
> commit c0bca60dc6977cca7352bacd1095081d89c88211
> Author: Shivank Garg <shivankg@amd.com>
> 
> dmaengine: fix use-after-free in dma_chan_put() and dma_release_channel()
> 
> When dma_device_put() drops the last reference on chan->device->ref,
> dma_device_release() may free the dma_device. This commit caches the
> module owner and delays dma_chan_put() to prevent reading freed memory.
> 
> Link: https://sashiko.dev/#/patchset/20260518-dmaengine-kref-fix-v1-1-4d6125048fb7@amd.com
> 
> > --- a/drivers/dma/dmaengine.c
> > +++ b/drivers/dma/dmaengine.c
> > @@ -916,8 +919,6 @@ void dma_release_channel(struct dma_chan *chan)
> >  	if (--chan->device->privatecnt == 0)
> >  		dma_cap_clear(DMA_PRIVATE, chan->device->cap_mask);
> >  
> > -	dma_chan_put(chan);
> > -
> >  	if (chan->slave) {
> >  		sysfs_remove_link(&chan->dev->device.kobj, DMA_SLAVE_NAME);
> >  		sysfs_remove_link(&chan->slave->kobj, chan->name);
> >  		kfree(chan->name);
> >  		chan->name = NULL;
> >  		chan->slave = NULL;
> >  	}
> >  
> >  #ifdef CONFIG_DEBUG_FS
> >  	kfree(chan->dbg_client_name);
> >  	chan->dbg_client_name = NULL;
> >  #endif
> > +	dma_chan_put(chan);
> 
> [Severity: High]
> Could moving dma_chan_put() here introduce a NULL pointer dereference
> during system suspend?
> 
> By moving dma_chan_put() after the block that clears chan->slave, the
> hardware channel may still be active while chan->slave is NULL. If a
> system suspend occurs concurrently, drivers like stm32-dma3 might check
> the hardware state and attempt to print the slave device name:
> 
> drivers/dma/stm32/stm32-dma3.c:stm32_dma3_pm_suspend() {
>     ...
> 	ccr = readl_relaxed(ddata->base + STM32_DMA3_CCR(chan->id));
> 	if (ccr & CCR_EN) {
> 		dev_warn(dev, "Suspend is prevented: %s still in use by %s\n",
> 			 dma_chan_name(c), dev_name(c->slave));
>     ...
> }
> 
> Since dma_chan_put() (which stops the hardware) hasn't been called yet,
> CCR_EN would still be active, and dev_name(c->slave) would dereference
> the NULL pointer we just set. 
> 
> Is there a way to stop the channel before clearing the slave pointer
> without triggering the original use-after-free?

This is a valid concern.
So, either we can have a NULL check in stm32 code while printing device
name, or drop dma_chan_put() change for now to avoid this trigger?

Thanks,
Shivank

[PATCH v3 3/3] dmaengine: wait for RCU readers before releasing dma_device

[Shivank Garg]

dma_issue_pending_all() walks the dma_device_list with
list_for_each_entry_rcu() under rcu_read_lock(). dma_device_release()
unlinks the device with list_del_rcu() and then calls
device->device_release() (which in many drivers, such as plx_dma.c,
directly calls kfree()).

Because there is no grace period between unlinking the device and
freeing it, concurrent RCU readers in dma_issue_pending_all() can
access the device after it has been freed.

The lockless walk originally relied on clients holding a dmaengine
reference to pin the provider module, and therefore the device, for as
long as they might traverse the list. Commit 8ad342a86359 ("dmaengine:
Add reference counting to dma_device struct") decoupled the dma_device
lifetime from the module reference, so the device can now be released
while a reader is still walking the list.

Add synchronize_rcu() before the device is freed, so RCU readers are
guaranteed to have finished. Keep it unconditional: providers that do
not implement device_release() free the device themselves once
dma_async_device_unregister() returns, so they need the same grace
period.

Fixes: 2ba05622b8b1 ("dmaengine: provide a common 'issue_pending_all' implementation")
Suggested-by: Sashiko <sashiko-bot@kernel.org>
Link: https://sashiko.dev/#/patchset/20260526-dmaengine-kref-fix-v2-0-3df60afac01d@amd.com
Signed-off-by: Shivank Garg <shivankg@amd.com>
---
 drivers/dma/dmaengine.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index bf491eb10596..29b128aa0bea 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -429,6 +429,12 @@ static void dma_device_release(struct kref *ref)
 	list_del_rcu(&device->global_node);
 	dma_channel_rebalance();
 
+	/*
+	 * Wait for RCU readers (e.g. dma_issue_pending_all()) that may still
+	 * be traversing dma_device_list before the device is freed.
+	 */
+	synchronize_rcu();
+
 	if (device->device_release)
 		device->device_release(device);
 }

-- 
2.43.0

Re: [PATCH v3 3/3] dmaengine: wait for RCU readers before releasing dma_device

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] Double list_del_rcu() causes list corruption when DMA provider module unloads concurrently with a channel request.
- [High] A Time-of-Check to Time-of-Use (TOCTOU) race condition between `dma_issue_pending_all()` and `dma_chan_put()` leads to memory corruption by operating on freed DMA channel resources.
--

--- Patch [3]: [PATCH v3 3/3] dmaengine: wait for RCU readers before releasing dma_device ---
commit 6738eea6e01509c5c24f2cdb086fe3fd0ffab13a
Author: Shivank Garg <shivankg@amd.com>

dmaengine: wait for RCU readers before releasing dma_device

dma_issue_pending_all() walks the dma_device_list with
list_for_each_entry_rcu() under rcu_read_lock(). dma_device_release()
unlinks the device with list_del_rcu() and then calls
device->device_release(). This patch adds synchronize_rcu() before
freeing the device so RCU readers are guaranteed to have finished.

Link: https://sashiko.dev/#/patchset/20260526-dmaengine-kref-fix-v2-0-3df60afac01d@amd.com

> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> index bf491eb10596d..29b128aa0beaa 100644
> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -429,6 +429,12 @@ static void dma_device_release(struct kref *ref)
>  	list_del_rcu(&device->global_node);

[Severity: High]
This isn't a bug introduced by this patch, but can this lead to a double
list_del_rcu() execution and list corruption if the module is unloaded
concurrently with a channel request?

If a thread calls dmaengine_get() or find_candidate(), it iterates
dma_device_list and calls dma_chan_get(). If another thread concurrently
unloads the module:

1. dma_chan_get() calls try_module_get(), which fails and returns -ENODEV
2. dmaengine_get() evaluates -ENODEV and eagerly deletes the device via
   list_del_rcu(&device->global_node)
3. Later, the unloading thread unregisters the device, which invokes
   dma_device_release()
4. dma_device_release() unconditionally calls
   list_del_rcu(&device->global_node) again on the already removed node

Calling list_del_rcu() twice on the same node poisons the prev pointer,
which can cause a panic in __list_del_entry().

>  	dma_channel_rebalance();
>  
> +	/*
> +	 * Wait for RCU readers (e.g. dma_issue_pending_all()) that may still
> +	 * be traversing dma_device_list before the device is freed.
> +	 */
> +	synchronize_rcu();

[Severity: High]
This is a pre-existing issue, but does dma_issue_pending_all() still have
a TOCTOU race condition with dma_chan_put() that could lead to memory
corruption?

While this synchronize_rcu() protects the dma_device itself during driver
unbind, it doesn't seem to protect against channel release.

If dma_issue_pending_all() evaluates chan->client_count locklessly
under rcu_read_lock():

Thread A:
dma_issue_pending_all()
    if (chan->client_count)
        // preempted here

Thread B:
dma_release_channel()
    dma_chan_put()
        // drops client_count to 0
        chan->device->device_free_chan_resources(chan);

Thread A resumes:
    device->device_issue_pending(chan);

The driver writes to the hardware, forcing it to fetch from the newly
freed DMA ring. Can the hardware be safely started here if the resources
were just freed?

> +
>  	if (device->device_release)
>  		device->device_release(device);
>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260816-dmaengine-kref-fix-v3-0-7e76187145df@amd.com?part=3