* [PATCH] misc: fastrpc: fix use-after-free in fastrpc_map_attach() error path
@ 2026-08-06 23:50 Yifei Gao
2026-08-07 10:04 ` Dmitry Baryshkov
2026-08-11 22:51 ` [PATCH v2] misc: fastrpc: fix double-free " Yifei Gao
0 siblings, 2 replies; 5+ messages in thread
From: Yifei Gao @ 2026-08-06 23:50 UTC (permalink / raw)
To: Srinivas Kandagatla, Amol Maheshwari
Cc: Greg Kroah-Hartman, Arnd Bergmann, Abel Vesa, Ekansh Gupta,
linux-arm-msm, dri-devel, linaro-mm-sig, linux-kernel, Yifei Gao,
stable
map->table is set right after the attachment is mapped, before the
len > map->size check. When that check fails and jumps to map_err, the
error path manually calls dma_buf_detach() and dma_buf_put(), then falls
through to fastrpc_map_put() -> fastrpc_free_map().
Since map->table is still non-NULL, fastrpc_free_map() repeats the
cleanup: dma_buf_unmap_attachment_unlocked() dereferences the map->attach
already freed by dma_buf_detach() (use-after-free read), and a second
dma_buf_put() drops an extra reference on map->buf. As the exporting fd is
typically still held by userspace, this imbalance can later lead to
premature destruction of the dma_buf and a use-after-free.
The branch is reachable by an unprivileged process via
FASTRPC_IOCTL_MEM_MAP with an fd whose dma-buf is smaller than the
requested length, before any DSP invocation.
Clear map->table in the map_err path so the fastrpc_map_put() fallthrough
does not operate on the already released attachment and buffer.
Fixes: 334f1a1cbe03 ("misc: fastrpc: Use fastrpc_map_put in fastrpc_map_create on fail")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-8
Signed-off-by: Yifei Gao <gyf161023@gmail.com>
---
drivers/misc/fastrpc.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
index f3a49384586d..d6be951d5538 100644
--- a/drivers/misc/fastrpc.c
+++ b/drivers/misc/fastrpc.c
@@ -915,6 +915,7 @@ static int fastrpc_map_attach(struct fastrpc_user *fl, int fd,
map_err:
dma_buf_detach(map->buf, map->attach);
+ map->table = NULL;
attach_err:
dma_buf_put(map->buf);
get_err:
--
2.43.0
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] misc: fastrpc: fix use-after-free in fastrpc_map_attach() error path 2026-08-06 23:50 [PATCH] misc: fastrpc: fix use-after-free in fastrpc_map_attach() error path Yifei Gao @ 2026-08-07 10:04 ` Dmitry Baryshkov 2026-08-11 22:51 ` [PATCH v2] misc: fastrpc: fix double-free " Yifei Gao 1 sibling, 0 replies; 5+ messages in thread From: Dmitry Baryshkov @ 2026-08-07 10:04 UTC (permalink / raw) To: Yifei Gao Cc: Srinivas Kandagatla, Amol Maheshwari, Greg Kroah-Hartman, Arnd Bergmann, Abel Vesa, Ekansh Gupta, linux-arm-msm, dri-devel, linaro-mm-sig, linux-kernel, stable On Thu, Aug 06, 2026 at 11:50:54PM +0000, Yifei Gao wrote: > map->table is set right after the attachment is mapped, before the > len > map->size check. When that check fails and jumps to map_err, the > error path manually calls dma_buf_detach() and dma_buf_put(), then falls > through to fastrpc_map_put() -> fastrpc_free_map(). > > Since map->table is still non-NULL, fastrpc_free_map() repeats the > cleanup: dma_buf_unmap_attachment_unlocked() dereferences the map->attach > already freed by dma_buf_detach() (use-after-free read), and a second > dma_buf_put() drops an extra reference on map->buf. As the exporting fd is > typically still held by userspace, this imbalance can later lead to > premature destruction of the dma_buf and a use-after-free. > > The branch is reachable by an unprivileged process via > FASTRPC_IOCTL_MEM_MAP with an fd whose dma-buf is smaller than the > requested length, before any DSP invocation. > > Clear map->table in the map_err path so the fastrpc_map_put() fallthrough > does not operate on the already released attachment and buffer. > > Fixes: 334f1a1cbe03 ("misc: fastrpc: Use fastrpc_map_put in fastrpc_map_create on fail") > Cc: stable@vger.kernel.org > Assisted-by: Claude:claude-opus-4-8 > Signed-off-by: Yifei Gao <gyf161023@gmail.com> > --- > drivers/misc/fastrpc.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c > index f3a49384586d..d6be951d5538 100644 > --- a/drivers/misc/fastrpc.c > +++ b/drivers/misc/fastrpc.c > @@ -915,6 +915,7 @@ static int fastrpc_map_attach(struct fastrpc_user *fl, int fd, > > map_err: > dma_buf_detach(map->buf, map->attach); > + map->table = NULL; This way it will skip dma_buf_unmap_attachment_unlocked() in fastrpc_map_free(), which is not nice. > attach_err: > dma_buf_put(map->buf); > get_err: > -- > 2.43.0 > -- With best wishes Dmitry ^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH v2] misc: fastrpc: fix double-free in fastrpc_map_attach() error path 2026-08-06 23:50 [PATCH] misc: fastrpc: fix use-after-free in fastrpc_map_attach() error path Yifei Gao 2026-08-07 10:04 ` Dmitry Baryshkov @ 2026-08-11 22:51 ` Yifei Gao 2026-08-20 4:44 ` Ekansh Gupta 2026-09-14 8:07 ` Srinivas Kandagatla 1 sibling, 2 replies; 5+ messages in thread From: Yifei Gao @ 2026-08-11 22:51 UTC (permalink / raw) To: Srinivas Kandagatla, Amol Maheshwari Cc: Greg Kroah-Hartman, Arnd Bergmann, Abel Vesa, Ekansh Gupta, Dmitry Baryshkov, linux-arm-msm, dri-devel, linaro-mm-sig, linux-kernel, Yifei Gao, stable map->table is assigned right after dma_buf_map_attachment_unlocked() succeeds. The two failure checks that follow, the len > map->size test and, where subsystem VMIDs are configured, a failed qcom_scm_assign_mem(), jump to map_err with map->table already set. map_err manually calls dma_buf_detach() and dma_buf_put() and then falls through to fastrpc_map_put(). Since that change the error path tail is fastrpc_map_put() -> fastrpc_free_map(), and fastrpc_free_map() already unmaps, detaches and puts the dma-buf whenever map->table is set. The two operations therefore run twice: the second dma_buf_put() drops an extra reference on map->buf, and dma_buf_unmap_attachment_unlocked() dereferences the map->attach already freed by the manual dma_buf_detach(). kref_init() sets the refcount to 1 with no intervening get, so the final fastrpc_map_put() frees the map synchronously and the redundant cleanup is deterministic. The len > map->size branch is reachable by an unprivileged process via FASTRPC_IOCTL_MEM_MAP with an fd whose dma-buf is smaller than the requested length, before any DSP invocation. Route both map->table-is-set failure branches to get_err instead of map_err, so fastrpc_free_map() is the single owner of the unmap/detach/put sequence. map_err is retained for the dma_buf_map_attachment_unlocked() failure, which is reached with map->table still NULL and an attachment that fastrpc_free_map() will not clean up, so its dma_buf_detach()/dma_buf_put() must still run manually. Fixes: 334f1a1cbe03 ("misc: fastrpc: Use fastrpc_map_put in fastrpc_map_create on fail") Cc: stable@vger.kernel.org Assisted-by: Claude:claude-opus-4-8 Signed-off-by: Yifei Gao <gyf161023@gmail.com> --- v2: - Instead of clearing map->table, route the map->table-is-set failure paths (len > map->size and the qcom_scm_assign_mem() failure) to get_err so fastrpc_free_map() is the single owner of the unmap/detach/put sequence. The map_err label is kept for the dma_buf_map_attachment_unlocked() failure, where map->table is still NULL. Suggested by Dmitry Baryshkov. drivers/misc/fastrpc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c index eb6c2a78d3c7..d480a87752a7 100644 --- a/drivers/misc/fastrpc.c +++ b/drivers/misc/fastrpc.c @@ -881,7 +881,7 @@ static int fastrpc_map_attach(struct fastrpc_user *fl, int fd, dev_dbg(sess->dev, "Bad size passed len 0x%llx map size 0x%llx\n", len, map->size); err = -EINVAL; - goto map_err; + goto get_err; } map->va = sg_virt(map->table->sgl); map->len = len; @@ -904,7 +904,7 @@ static int fastrpc_map_attach(struct fastrpc_user *fl, int fd, dev_err(sess->dev, "Failed to assign memory with dma_addr %pad size 0x%llx err %d\n", &map->dma_addr, map->len, err); - goto map_err; + goto get_err; } } spin_lock(&fl->lock); -- 2.43.0 ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH v2] misc: fastrpc: fix double-free in fastrpc_map_attach() error path 2026-08-11 22:51 ` [PATCH v2] misc: fastrpc: fix double-free " Yifei Gao @ 2026-08-20 4:44 ` Ekansh Gupta 2026-09-14 8:07 ` Srinivas Kandagatla 1 sibling, 0 replies; 5+ messages in thread From: Ekansh Gupta @ 2026-08-20 4:44 UTC (permalink / raw) To: Yifei Gao, Srinivas Kandagatla, Amol Maheshwari Cc: Greg Kroah-Hartman, Arnd Bergmann, Abel Vesa, Dmitry Baryshkov, linux-arm-msm, dri-devel, linaro-mm-sig, linux-kernel, stable On 12-08-2026 04:21, Yifei Gao wrote: > map->table is assigned right after dma_buf_map_attachment_unlocked() > succeeds. The two failure checks that follow, the len > map->size test > and, where subsystem VMIDs are configured, a failed qcom_scm_assign_mem(), > jump to map_err with map->table already set. > > map_err manually calls dma_buf_detach() and dma_buf_put() and then falls > through to fastrpc_map_put(). Since that change the error path tail is > fastrpc_map_put() -> fastrpc_free_map(), and fastrpc_free_map() already > unmaps, detaches and puts the dma-buf whenever map->table is set. > The two operations therefore run twice: the second dma_buf_put() drops an > extra reference on map->buf, and dma_buf_unmap_attachment_unlocked() > dereferences the map->attach already freed by the manual dma_buf_detach(). > kref_init() sets the refcount to 1 with no intervening get, so the final > fastrpc_map_put() frees the map synchronously and the redundant cleanup is > deterministic. > > The len > map->size branch is reachable by an unprivileged process via > FASTRPC_IOCTL_MEM_MAP with an fd whose dma-buf is smaller than the > requested length, before any DSP invocation. > > Route both map->table-is-set failure branches to get_err instead of > map_err, so fastrpc_free_map() is the single owner of the > unmap/detach/put sequence. map_err is retained for the > dma_buf_map_attachment_unlocked() failure, which is reached with > map->table still NULL and an attachment that fastrpc_free_map() will not > clean up, so its dma_buf_detach()/dma_buf_put() must still run manually. > > Fixes: 334f1a1cbe03 ("misc: fastrpc: Use fastrpc_map_put in fastrpc_map_create on fail") > Cc: stable@vger.kernel.org > Assisted-by: Claude:claude-opus-4-8 > Signed-off-by: Yifei Gao <gyf161023@gmail.com> Reviewed-by: Ekansh Gupta <ekansh.gupta@oss.qualcomm.com> ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH v2] misc: fastrpc: fix double-free in fastrpc_map_attach() error path 2026-08-11 22:51 ` [PATCH v2] misc: fastrpc: fix double-free " Yifei Gao 2026-08-20 4:44 ` Ekansh Gupta @ 2026-09-14 8:07 ` Srinivas Kandagatla 1 sibling, 0 replies; 5+ messages in thread From: Srinivas Kandagatla @ 2026-09-14 8:07 UTC (permalink / raw) To: Amol Maheshwari, Yifei Gao Cc: Greg Kroah-Hartman, Arnd Bergmann, Abel Vesa, Ekansh Gupta, Dmitry Baryshkov, linux-arm-msm, dri-devel, linaro-mm-sig, linux-kernel, stable On Tue, 11 Aug 2026 22:51:01 +0000, Yifei Gao wrote: > map->table is assigned right after dma_buf_map_attachment_unlocked() > succeeds. The two failure checks that follow, the len > map->size test > and, where subsystem VMIDs are configured, a failed qcom_scm_assign_mem(), > jump to map_err with map->table already set. > > map_err manually calls dma_buf_detach() and dma_buf_put() and then falls > through to fastrpc_map_put(). Since that change the error path tail is > fastrpc_map_put() -> fastrpc_free_map(), and fastrpc_free_map() already > unmaps, detaches and puts the dma-buf whenever map->table is set. > The two operations therefore run twice: the second dma_buf_put() drops an > extra reference on map->buf, and dma_buf_unmap_attachment_unlocked() > dereferences the map->attach already freed by the manual dma_buf_detach(). > kref_init() sets the refcount to 1 with no intervening get, so the final > fastrpc_map_put() frees the map synchronously and the redundant cleanup is > deterministic. > > [...] Applied, thanks! [1/1] misc: fastrpc: fix double-free in fastrpc_map_attach() error path commit: 2159430fe26068ca3c89557ddd2d5eb17a5a8cf0 Best regards, -- Srinivas Kandagatla <srini@kernel.org> ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2026-09-14 8:07 UTC | newest] Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2026-08-06 23:50 [PATCH] misc: fastrpc: fix use-after-free in fastrpc_map_attach() error path Yifei Gao 2026-08-07 10:04 ` Dmitry Baryshkov 2026-08-11 22:51 ` [PATCH v2] misc: fastrpc: fix double-free " Yifei Gao 2026-08-20 4:44 ` Ekansh Gupta 2026-09-14 8:07 ` Srinivas Kandagatla
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®