drivers/atm/fore200e.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
When the PCA-200E or SBA-200E adapter is being detached, the fore200e
is deallocated. However, the tx_tasklet or rx_tasklet may still be running
or pending, leading to use-after-free bug when the already freed fore200e
is accessed again in fore200e_tx_tasklet() or fore200e_rx_tasklet().
One of the race conditions can occur as follows:
CPU 0 (cleanup) | CPU 1 (tasklet)
fore200e_pca_remove_one() | fore200e_interrupt()
fore200e_shutdown() | tasklet_schedule()
kfree(fore200e) | fore200e_tx_tasklet()
| fore200e-> // UAF
Fix this by ensuring tx_tasklet or rx_tasklet is properly canceled before
the fore200e is released. Add tasklet_kill() in fore200e_shutdown() to
synchronize with any pending or running tasklets. Moreover, since
fore200e_reset() could prevent further interrupts or data transfers,
the tasklet_kill() should be placed after fore200e_reset() to prevent
the tasklet from being rescheduled in fore200e_interrupt(). Finally,
it only needs to do tasklet_kill() when the fore200e state is greater
than or equal to FORE200E_STATE_IRQ, since tasklets are uninitialized
in earlier states.
This bug was identified through static analysis.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Cc: stable@kernel.org
Suggested-by: Jijie Shao <shaojijie@huawei.com>
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
---
Changes in v3:
- Only do tasklet_kill() when state >= FORE200E_STATE_IRQ.
- Make the commit messages clearer.
drivers/atm/fore200e.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/atm/fore200e.c b/drivers/atm/fore200e.c
index f62e3857144..eb386ec6b5e 100644
--- a/drivers/atm/fore200e.c
+++ b/drivers/atm/fore200e.c
@@ -363,7 +363,12 @@ fore200e_shutdown(struct fore200e* fore200e)
/* first, reset the board to prevent further interrupts or data transfers */
fore200e_reset(fore200e, 0);
}
-
+#ifdef FORE200E_USE_TASKLET
+ if (fore200e->state >= FORE200E_STATE_IRQ) {
+ tasklet_kill(&fore200e->tx_tasklet);
+ tasklet_kill(&fore200e->rx_tasklet);
+ }
+#endif
/* then, release all allocated resources */
switch(fore200e->state) {
--
2.34.1Hi Duoming,
kernel test robot noticed the following build warnings:
[auto build test WARNING on net/main]
url: https://github.com/intel-lab-lkp/linux/commits/Duoming-Zhou/atm-fore200e-fix-use-after-free-in-tasklets-during-device-removal/20260209-174706
base: net/main
patch link: https://lore.kernel.org/r/20260209094512.33847-1-duoming%40zju.edu.cn
patch subject: [PATCH net v3] atm: fore200e: fix use-after-free in tasklets during device removal
config: i386-randconfig-141-20260210 (https://download.01.org/0day-ci/archive/20260210/202602101146.NCGz3JHc-lkp@intel.com/config)
compiler: clang version 20.1.8 (https://github.com/llvm/llvm-project 87f0227cb60147a26a1eeb4fb06e3b505e9c7261)
smatch version: v0.5.0-8994-gd50c5a4c
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202602101146.NCGz3JHc-lkp@intel.com/
smatch warnings:
drivers/atm/fore200e.c:367 fore200e_shutdown() warn: inconsistent indenting
vim +367 drivers/atm/fore200e.c
353
354
355 static void
356 fore200e_shutdown(struct fore200e* fore200e)
357 {
358 printk(FORE200E "removing device %s at 0x%lx, IRQ %s\n",
359 fore200e->name, fore200e->phys_base,
360 fore200e_irq_itoa(fore200e->irq));
361
362 if (fore200e->state > FORE200E_STATE_RESET) {
363 /* first, reset the board to prevent further interrupts or data transfers */
364 fore200e_reset(fore200e, 0);
365 }
366 #ifdef FORE200E_USE_TASKLET
> 367 if (fore200e->state >= FORE200E_STATE_IRQ) {
368 tasklet_kill(&fore200e->tx_tasklet);
369 tasklet_kill(&fore200e->rx_tasklet);
370 }
371 #endif
372 /* then, release all allocated resources */
373 switch(fore200e->state) {
374
375 case FORE200E_STATE_COMPLETE:
376 kfree(fore200e->stats);
377
378 fallthrough;
379 case FORE200E_STATE_IRQ:
380 free_irq(fore200e->irq, fore200e->atm_dev);
381
382 fallthrough;
383 case FORE200E_STATE_ALLOC_BUF:
384 fore200e_free_rx_buf(fore200e);
385
386 fallthrough;
387 case FORE200E_STATE_INIT_BSQ:
388 fore200e_uninit_bs_queue(fore200e);
389
390 fallthrough;
391 case FORE200E_STATE_INIT_RXQ:
392 fore200e_dma_chunk_free(fore200e, &fore200e->host_rxq.status);
393 fore200e_dma_chunk_free(fore200e, &fore200e->host_rxq.rpd);
394
395 fallthrough;
396 case FORE200E_STATE_INIT_TXQ:
397 fore200e_dma_chunk_free(fore200e, &fore200e->host_txq.status);
398 fore200e_dma_chunk_free(fore200e, &fore200e->host_txq.tpd);
399
400 fallthrough;
401 case FORE200E_STATE_INIT_CMDQ:
402 fore200e_dma_chunk_free(fore200e, &fore200e->host_cmdq.status);
403
404 fallthrough;
405 case FORE200E_STATE_INITIALIZE:
406 /* nothing to do for that state */
407
408 case FORE200E_STATE_START_FW:
409 /* nothing to do for that state */
410
411 case FORE200E_STATE_RESET:
412 /* nothing to do for that state */
413
414 case FORE200E_STATE_MAP:
415 fore200e->bus->unmap(fore200e);
416
417 fallthrough;
418 case FORE200E_STATE_CONFIGURE:
419 /* nothing to do for that state */
420
421 case FORE200E_STATE_REGISTER:
422 /* XXX shouldn't we *start* by deregistering the device? */
423 atm_dev_deregister(fore200e->atm_dev);
424
425 fallthrough;
426 case FORE200E_STATE_BLANK:
427 /* nothing to do for that state */
428 break;
429 }
430 }
431
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
on 2026/2/9 17:45, Duoming Zhou wrote:
> When the PCA-200E or SBA-200E adapter is being detached, the fore200e
> is deallocated. However, the tx_tasklet or rx_tasklet may still be running
> or pending, leading to use-after-free bug when the already freed fore200e
> is accessed again in fore200e_tx_tasklet() or fore200e_rx_tasklet().
>
> One of the race conditions can occur as follows:
>
> CPU 0 (cleanup) | CPU 1 (tasklet)
> fore200e_pca_remove_one() | fore200e_interrupt()
> fore200e_shutdown() | tasklet_schedule()
> kfree(fore200e) | fore200e_tx_tasklet()
> | fore200e-> // UAF
>
> Fix this by ensuring tx_tasklet or rx_tasklet is properly canceled before
> the fore200e is released. Add tasklet_kill() in fore200e_shutdown() to
> synchronize with any pending or running tasklets. Moreover, since
> fore200e_reset() could prevent further interrupts or data transfers,
> the tasklet_kill() should be placed after fore200e_reset() to prevent
> the tasklet from being rescheduled in fore200e_interrupt(). Finally,
> it only needs to do tasklet_kill() when the fore200e state is greater
> than or equal to FORE200E_STATE_IRQ, since tasklets are uninitialized
> in earlier states.
>
> This bug was identified through static analysis.
>
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Cc: stable@kernel.org
> Suggested-by: Jijie Shao <shaojijie@huawei.com>
> Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> ---
> Changes in v3:
> - Only do tasklet_kill() when state >= FORE200E_STATE_IRQ.
> - Make the commit messages clearer.
>
> drivers/atm/fore200e.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/atm/fore200e.c b/drivers/atm/fore200e.c
> index f62e3857144..eb386ec6b5e 100644
> --- a/drivers/atm/fore200e.c
> +++ b/drivers/atm/fore200e.c
> @@ -363,7 +363,12 @@ fore200e_shutdown(struct fore200e* fore200e)
> /* first, reset the board to prevent further interrupts or data transfers */
> fore200e_reset(fore200e, 0);
> }
> -
> +#ifdef FORE200E_USE_TASKLET
> + if (fore200e->state >= FORE200E_STATE_IRQ) {
> + tasklet_kill(&fore200e->tx_tasklet);
> + tasklet_kill(&fore200e->rx_tasklet);
> + }
> +#endif
The code appears to be misaligned.
I see that in the original driver code, the first level of indentation uses 4 spaces,
and the second level uses a TAB (8-character width).
In your code, the first level uses a TAB, and the second level uses two TABs.
Of course, your code complies with the current coding standards.
Additionally, I think a better modification would be to do tasklet_kill()
in the "switch(fore200e->state)-case FORE200E_STATE_IRQ" branch,
because "fore200e->tx_tasklet" is initialized together with "fore200e->irq".
Of course, your current change also looks fine, so
Reviewed-by: Jijie Shao <shaojijie@huawei.com>
On Tue, 10 Feb 2026 09:46:53 +0800 Jijie Shao wrote:
> > When the PCA-200E or SBA-200E adapter is being detached, the fore200e
> > is deallocated. However, the tx_tasklet or rx_tasklet may still be running
> > or pending, leading to use-after-free bug when the already freed fore200e
> > is accessed again in fore200e_tx_tasklet() or fore200e_rx_tasklet().
> >
> > One of the race conditions can occur as follows:
> >
> > CPU 0 (cleanup) | CPU 1 (tasklet)
> > fore200e_pca_remove_one() | fore200e_interrupt()
> > fore200e_shutdown() | tasklet_schedule()
> > kfree(fore200e) | fore200e_tx_tasklet()
> > | fore200e-> // UAF
> >
> > Fix this by ensuring tx_tasklet or rx_tasklet is properly canceled before
> > the fore200e is released. Add tasklet_kill() in fore200e_shutdown() to
> > synchronize with any pending or running tasklets. Moreover, since
> > fore200e_reset() could prevent further interrupts or data transfers,
> > the tasklet_kill() should be placed after fore200e_reset() to prevent
> > the tasklet from being rescheduled in fore200e_interrupt(). Finally,
> > it only needs to do tasklet_kill() when the fore200e state is greater
> > than or equal to FORE200E_STATE_IRQ, since tasklets are uninitialized
> > in earlier states.
> >
> > This bug was identified through static analysis.
> >
> > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> > Cc: stable@kernel.org
> > Suggested-by: Jijie Shao <shaojijie@huawei.com>
> > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > ---
> > Changes in v3:
> > - Only do tasklet_kill() when state >= FORE200E_STATE_IRQ.
> > - Make the commit messages clearer.
> >
> > drivers/atm/fore200e.c | 7 ++++++-
> > 1 file changed, 6 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/atm/fore200e.c b/drivers/atm/fore200e.c
> > index f62e3857144..eb386ec6b5e 100644
> > --- a/drivers/atm/fore200e.c
> > +++ b/drivers/atm/fore200e.c
> > @@ -363,7 +363,12 @@ fore200e_shutdown(struct fore200e* fore200e)
> > /* first, reset the board to prevent further interrupts or data transfers */
> > fore200e_reset(fore200e, 0);
> > }
> > -
> > +#ifdef FORE200E_USE_TASKLET
> > + if (fore200e->state >= FORE200E_STATE_IRQ) {
> > + tasklet_kill(&fore200e->tx_tasklet);
> > + tasklet_kill(&fore200e->rx_tasklet);
> > + }
> > +#endif
>
> The code appears to be misaligned.
> I see that in the original driver code, the first level of indentation uses 4 spaces,
> and the second level uses a TAB (8-character width).
> In your code, the first level uses a TAB, and the second level uses two TABs.
> Of course, your code complies with the current coding standards.
>
> Additionally, I think a better modification would be to do tasklet_kill()
> in the "switch(fore200e->state)-case FORE200E_STATE_IRQ" branch,
> because "fore200e->tx_tasklet" is initialized together with "fore200e->irq".
Thank you for your suggestions! I will adjust the tasklet_kill() into the
"case FORE200E_STATE_IRQ" branch. The details are as follows:
diff --git a/drivers/atm/fore200e.c b/drivers/atm/fore200e.c
index f62e3857144..fec081db36d 100644
--- a/drivers/atm/fore200e.c
+++ b/drivers/atm/fore200e.c
@@ -373,6 +373,10 @@ fore200e_shutdown(struct fore200e* fore200e)
fallthrough;
case FORE200E_STATE_IRQ:
free_irq(fore200e->irq, fore200e->atm_dev);
+#ifdef FORE200E_USE_TASKLET
+ tasklet_kill(&fore200e->tx_tasklet);
+ tasklet_kill(&fore200e->rx_tasklet);
+#endif
Best regards,
Duoming Zhou
© 2016 - 2026 Red Hat, Inc.