net/9p/usbg: fix descriptor cleanup use-after-free
Subsystems (2)
PATCH 1/1 net/9p/usbg: fix descriptor cleanup use-after-free
usb9pfs_free_func() frees the f_usb9pfs object before calling usb_free_all_descriptors(). The usb_function passed to the latter is embedded in the freed object, so removing the function from a gadget dereferences freed memory.
The failing teardown sequence is:
configfs unlink
usb9pfs_free_func()
kfree(usb9pfs)
usb_free_all_descriptors(f)
dereference the embedded usb_function
KASAN reported:
BUG: KASAN: slab-use-after-free in usb_free_all_descriptors+0x138/0x190 Read of size 8 at addr ffff888106a73088 by task poc/95 Call Trace: usb_free_all_descriptors+0x138/0x190 config_usb_cfg_unlink+0x1f0/0x2f0 configfs_unlink+0x321/0x6f0
Free the descriptors before freeing their containing object.
--- net/9p/trans_usbg.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
net/9p/trans_usbg.c +1 -2
@@ -725,8 +725,6 @@ static void usb9pfs_free_func(struct usb_function *f)struct f_usb9pfs *usb9pfs =func_to_usb9pfs (f);struct f_usb9pfs_opts *opts; -kfree (usb9pfs); - opts = container_of(f->fi, struct f_usb9pfs_opts, func_inst);mutex_lock (&opts->lock);
@@ -734,6 +732,7 @@ static void usb9pfs_free_func(struct usb_function *f)mutex_unlock (&opts->lock);usb_free_all_descriptors (f); +kfree (usb9pfs); } static int usb9pfs_set_alt(struct usb_function *f,
Quoted 15 lines · Chengfeng Ye
usb9pfs_free_func()frees thef_usb9pfsobject before callingusb_free_all_descriptors(). Theusb_functionpassed to the latter is embedded in the freed object, so removing the function from a gadget dereferences freed memory.The failing teardown sequence is:
configfs unlink
usb9pfs_free_func()
kfree(usb9pfs)
usb_free_all_descriptors(f)
dereference the embeddedusb_functionKASAN reported:
Quoted 8 lines
BUG: KASAN: slab-use-after-free in usb_free_all_descriptors+0x138/0x190 Read of size 8 at addr ffff888106a73088 by task poc/95 Call Trace: usb_free_all_descriptors+0x138/0x190 config_usb_cfg_unlink+0x1f0/0x2f0 configfs_unlink+0x321/0x6f0Free the descriptors before freeing their containing object.
Although this is a copy of https://lore.kernel.org/all/[email protected]/
which actually came earlier than your, I like the commit message here better.
Quoted 28 lines
Fixes a3be076dc174 ("net/9p/usbg: Add new usb gadget function transport")Signed-off-by Chengfeng Ye <[email protected]>--- net/9p/trans_usbg.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)net/9p/trans_usbg.c +1
-2@@ -725,8 +725,6 @@ static void usb9pfs_free_func(struct usb_function *f)struct f_usb9pfs *usb9pfs =func_to_usb9pfs (f);struct f_usb9pfs_opts *opts; -kfree (usb9pfs); - opts = container_of(f->fi, struct f_usb9pfs_opts, func_inst);mutex_lock (&opts->lock);@@ -734,6 +732,7 @@ static void usb9pfs_free_func(struct usb_function *f)mutex_unlock (&opts->lock);usb_free_all_descriptors (f); +kfree (usb9pfs); } static int usb9pfs_set_alt(struct usb_function *f,
Dominique Martinet
maintainerChengfeng Ye wrote on Fri, Jul 31, 2026 at 11:04:14PM +0800:
Quoted 22 lines
usb9pfs_free_func()frees thef_usb9pfsobject before callingusb_free_all_descriptors(). Theusb_functionpassed to the latter is embedded in the freed object, so removing the function from a gadget dereferences freed memory.The failing teardown sequence is:
configfs unlink
usb9pfs_free_func()
kfree(usb9pfs)
usb_free_all_descriptors(f)
dereference the embeddedusb_functionKASAN reported:
BUG: KASAN: slab-use-after-free in usb_free_all_descriptors+0x138/0x190 Read of size 8 at addr ffff888106a73088 by task poc/95 Call Trace: usb_free_all_descriptors+0x138/0x190 config_usb_cfg_unlink+0x1f0/0x2f0 configfs_unlink+0x321/0x6f0
Quoted 5 lines
Free the descriptors before freeing their containing object.
Fixes a3be076dc174 ("net/9p/usbg: Add new usb gadget function transport")Signed-off-by Chengfeng Ye <[email protected]>
Thanks for the patch, given there was another identical patch earlier but Michael preferred your commit message I've kept the original patch with your commit message, keeping you as Co-developed-by
It should get in 7.4 if I'm not hit by a bus
other mail
https://lore.kernel.org/r/[email protected]
commit as of now
https://github.com/martinetd/linux/commit/0955472fff1e9f9e410a0300136924ebc8bd7a16
(look in https://github.com/martinetd/linux/commits/9p-test if commit is gone on rebase)
Signature · 1 line
Dominique Martinet | Asmadeus