Building the Linux kernel with Clang and LLVM
 help / color / mirror / Atom feed
From: Rosen Penev <rosenp@gmail.com>
To: dmaengine@vger.kernel.org
Cc: Vinod Koul <vkoul@kernel.org>, Frank Li <Frank.Li@kernel.org>,
	Zhang Wei <zw@zh-kernel.org>,
	Nathan Chancellor <nathan@kernel.org>,
	Nick Desaulniers <nick.desaulniers+lkml@gmail.com>,
	Bill Wendling <morbo@google.com>,
	Justin Stitt <justinstitt@google.com>,
	linux-kernel@vger.kernel.org (open list),
	linuxppc-dev@lists.ozlabs.org (open list:FREESCALE DMA DRIVER),
	llvm@lists.linux.dev (open list:CLANG/LLVM BUILD
	SUPPORT:Keyword:\b(?i:clang|llvm)\b)
Subject: [PATCHv3 04/15] dmaengine: fsldma: provide device_release callback
Date: Tue,  9 Jun 2026 15:19:15 -0700	[thread overview]
Message-ID: <20260609221926.35538-5-rosenp@gmail.com> (raw)
In-Reply-To: <20260609221926.35538-1-rosenp@gmail.com>

The DMA core requires drivers to set dma_device.device_release so that
the container structure is only freed after all references to it have
been dropped (see the comment above dma_async_device_register()).

This driver violated that contract: fdev was devm_kzalloc()'d with no
device_release callback.  If a client still held a channel reference
when the driver was unbound, dma_device_release() would eventually
run on freed memory, causing a use-after-free.

Fix by allocating fdev with kzalloc_obj(), adding
fsldma_device_release() to free it, and setting device_release.
fsldma_of_remove() now saves channel pointers and frees IRQs before
calling dma_async_device_unregister(), since fdev may be freed by
the release callback inside that call.

Assisted-by: opencode:big-pickle
Signed-off-by: Rosen Penev <rosenp@gmail.com>
---
 drivers/dma/fsldma.c | 27 ++++++++++++++++++++++-----
 1 file changed, 22 insertions(+), 5 deletions(-)

diff --git a/drivers/dma/fsldma.c b/drivers/dma/fsldma.c
index 1ba10d065278..43d817f6ded1 100644
--- a/drivers/dma/fsldma.c
+++ b/drivers/dma/fsldma.c
@@ -1219,6 +1219,8 @@ static void fsl_dma_chan_remove(struct fsldma_chan *chan)
 	kfree(chan);
 }
 
+static void fsldma_device_release(struct dma_device *dma_dev);
+
 static int fsldma_of_probe(struct platform_device *op)
 {
 	struct fsldma_device *fdev;
@@ -1257,6 +1259,7 @@ static int fsldma_of_probe(struct platform_device *op)
 	fdev->common.device_issue_pending = fsl_dma_memcpy_issue_pending;
 	fdev->common.device_config = fsl_dma_device_config;
 	fdev->common.device_terminate_all = fsl_dma_device_terminate_all;
+	fdev->common.device_release = fsldma_device_release;
 	fdev->common.dev = &op->dev;
 
 	fdev->common.src_addr_widths = FSL_DMA_BUSWIDTHS;
@@ -1316,19 +1319,33 @@ static int fsldma_of_probe(struct platform_device *op)
 	return err;
 }
 
+static void fsldma_device_release(struct dma_device *dma_dev)
+{
+	struct fsldma_device *fdev = container_of(dma_dev, struct fsldma_device,
+						  common);
+	kfree(fdev);
+}
+
 static void fsldma_of_remove(struct platform_device *op)
 {
-	struct fsldma_device *fdev;
+	struct fsldma_device *fdev = platform_get_drvdata(op);
+	struct fsldma_chan *chans[FSL_DMA_MAX_CHANS_PER_DEVICE];
 	unsigned int i;
 
-	fdev = platform_get_drvdata(op);
-	dma_async_device_unregister(&fdev->common);
+	for (i = 0; i < FSL_DMA_MAX_CHANS_PER_DEVICE; i++)
+		chans[i] = fdev->chan[i];
 
 	fsldma_free_irqs(fdev);
 
+	/*
+	 * fdev may be freed by fsldma_device_release inside this call;
+	 * use saved copies of the channel pointers afterwards.
+	 */
+	dma_async_device_unregister(&fdev->common);
+
 	for (i = 0; i < FSL_DMA_MAX_CHANS_PER_DEVICE; i++) {
-		if (fdev->chan[i])
-			fsl_dma_chan_remove(fdev->chan[i]);
+		if (chans[i])
+			fsl_dma_chan_remove(chans[i]);
 	}
 	irq_dispose_mapping(fdev->irq);
 
-- 
2.54.0


  parent reply	other threads:[~2026-06-09 22:19 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-06-09 22:19 [PATCHv3 00/15] dmaengine: fsldma: devm conversion, fixups, and cleanups Rosen Penev
2026-06-09 22:19 ` [PATCHv3 01/15] dmaengine: fsldma: kill tasklet before removing channel Rosen Penev
2026-06-10  1:35   ` Frank Li
2026-06-09 22:19 ` [PATCHv3 02/15] dmaengine: fsldma: drop desc_lock before invoking client callback Rosen Penev
2026-06-09 22:19 ` [PATCHv3 03/15] dmaengine: fsldma: halt DMA engine before freeing resources Rosen Penev
2026-06-10  2:46   ` Frank Li
2026-06-09 22:19 ` Rosen Penev [this message]
2026-06-09 22:19 ` [PATCHv3 05/15] dmaengine: fsldma: check dma_async_device_register() return value Rosen Penev
2026-06-09 22:19 ` [PATCHv3 06/15] dmaengine: fsldma: fix probe error path not freeing IRQs Rosen Penev
2026-06-09 22:19 ` [PATCHv3 07/15] dmaengine: fsldma: fix request_irqs unwind freeing unregistered IRQ Rosen Penev
2026-06-09 22:19 ` [PATCHv3 08/15] dmaengine: fsldma: convert to platform_get_irq_optional() Rosen Penev
2026-06-10  2:58   ` Frank Li
2026-06-09 22:19 ` [PATCHv3 09/15] dmaengine: fsldma: use devm for kzalloc() Rosen Penev
2026-06-10  1:57   ` Frank Li
2026-06-09 22:19 ` [PATCHv3 10/15] dmaengine: fsldma: use devm_platform_ioremap_resource() Rosen Penev
2026-06-09 22:19 ` [PATCHv3 11/15] dmaengine: fsldma: convert channel allocation to devm_kzalloc() Rosen Penev
2026-06-09 22:19 ` [PATCHv3 12/15] dmaengine: fsldma: use devm for of_iomap() Rosen Penev
2026-06-10  1:53   ` Frank Li
2026-06-09 22:19 ` [PATCHv3 13/15] dmaengine: fsldma: replace irq_of_parse_and_map with of_irq_get Rosen Penev
2026-06-09 22:19 ` [PATCHv3 14/15] dmaengine: fsldma: replace ppc-specific accessors with portable generic ones Rosen Penev
2026-06-09 22:19 ` [PATCHv3 15/15] dmaengine: fsldma: fix kernel-doc param names to match function signatures Rosen Penev

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=20260609221926.35538-5-rosenp@gmail.com \
    --to=rosenp@gmail.com \
    --cc=Frank.Li@kernel.org \
    --cc=dmaengine@vger.kernel.org \
    --cc=justinstitt@google.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linuxppc-dev@lists.ozlabs.org \
    --cc=llvm@lists.linux.dev \
    --cc=morbo@google.com \
    --cc=nathan@kernel.org \
    --cc=nick.desaulniers+lkml@gmail.com \
    --cc=vkoul@kernel.org \
    --cc=zw@zh-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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox