9p/fd: set req refcount to zero to avoid uninitialized usage
Subsystems (3)
PATCH v3 1/1 9p/fd: set req refcount to zero to avoid uninitialized usage
When the new request allocated, the refcount will be zero if it is resued one. But if the request is newly allocated from slab, it is not fully initialized before add it to idr.
If the p9_read_work got a response before the refcount initiated. It will use a uninitialized req, which will result in a bad request data struct.
Here is the logs from syzbot.
Corrupted memory at 0xffff88807eade00b [ 0xff 0x07 0x00 0x00 0x00 0x00 0x00 0x00 . . . . . . . . ] (in kfence-#110):
p9_fcall_fini net/9p/client.c:248 [inline] p9_req_put net/9p/client.c:396 [inline] p9_req_put+0x208/0x250 net/9p/client.c:390 p9_client_walk+0x247/0x540 net/9p/client.c:1165 clone_fid fs/9p/fid.h:21 [inline] v9fs_fid_xattr_set+0xe4/0x2b0 fs/9p/xattr.c:118 v9fs_xattr_set fs/9p/xattr.c:100 [inline] v9fs_xattr_handler_set+0x6f/0x120 fs/9p/xattr.c:159 __vfs_setxattr+0x119/0x180 fs/xattr.c:182 __vfs_setxattr_noperm+0x129/0x5f0 fs/xattr.c:216 __vfs_setxattr_locked+0x1d3/0x260 fs/xattr.c:277 vfs_setxattr+0x143/0x340 fs/xattr.c:309 setxattr+0x146/0x160 fs/xattr.c:617 path_setxattr+0x197/0x1c0 fs/xattr.c:636 __do_sys_setxattr fs/xattr.c:652 [inline] __se_sys_setxattr fs/xattr.c:648 [inline] __ia32_sys_setxattr+0xc0/0x160 fs/xattr.c:648 do_syscall_32_irqs_on arch/x86/entry/common.c:112 [inline] __do_fast_syscall_32+0x65/0xf0 arch/x86/entry/common.c:178 do_fast_syscall_32+0x33/0x70 arch/x86/entry/common.c:203 entry_SYSENTER_compat_after_hwframe+0x70/0x82
Below is a similar scenario, the scenario in the syzbot log looks more complicated than this one, but this patch can fix it.
T21124 p9_read_work
======================== second trans ================================= p9_client_walk
p9_client_rpc
p9_client_prepare_req
p9_tag_alloc
req = kmem_cache_alloc(p9_req_cache, GFP_NOFS);
tag = idr_alloc
<< preempted >>
req->tc.tag = tag;
/* req->[refcount/tag] == uninitialized */
m->rreq = p9_tag_lookup(m->client, m->rc.tag);
/* increments uninitalized refcount */
refcount_set(&req->refcount, 2);
/* cb drops one ref */
p9_client_cb(req)
/* reader thread drops its ref:
request is incorrectly freed */
p9_req_put(req)
/* use after free and ref underflow */
p9_req_put(req)To fix it, we can initize the refcount to zero before add to idr.
--
Changelog:
v1 -> v2:
- Set refcount to fix the problem.
v2 -> v3:
- Comment messages improve as asmadeus suggested.
--- net/9p/client.c | 5 +++++ 1 file changed, 5 insertions(+)
net/9p/client.c +5 -0
@@ -297,6 +297,11 @@ p9_tag_alloc(struct p9_client *c, int8_t type, uint t_size, uint r_size,p9pdu_reset (&req->rc); req->t_err =0 ; req->status =REQ_STATUS_ALLOC ; +/* refcount needs to be set to 0 before inserting into the idr +* so p9_tag_lookup does not accept a request that is not fully +* initialized. refcount_set to 2 below will mark request live. +*/ +refcount_set (&req->refcount,0 );init_waitqueue_head (&req->wq);INIT_LIST_HEAD (&req->req_list);
Signature · 1 line
2.37.3
Quoted 59 lines · CET Schspa Shi
When the new request allocated, the refcount will be zero if it is resued one. But if the request is newly allocated from slab, it is not fully initialized before add it to idr.
If the
p9_read_workgot a response before the refcount initiated. It will use a uninitialized req, which will result in a bad request data struct.Here is the logs from syzbot.
Corrupted memory at
0xffff88807eade00b[0xff0x070x000x000x000x000x000x00. . . . . . . . ] (in kfence-#110):p9_fcall_fini net/9p/client.c:248 [inline] p9_req_put net/9p/client.c:396 [inline] p9_req_put+0x208/0x250 net/9p/client.c:390 p9_client_walk+0x247/0x540 net/9p/client.c:1165 clone_fid fs/9p/fid.h:21 [inline] v9fs_fid_xattr_set+0xe4/0x2b0 fs/9p/xattr.c:118 v9fs_xattr_set fs/9p/xattr.c:100 [inline] v9fs_xattr_handler_set+0x6f/0x120 fs/9p/xattr.c:159 __vfs_setxattr+0x119/0x180 fs/xattr.c:182 __vfs_setxattr_noperm+0x129/0x5f0 fs/xattr.c:216 __vfs_setxattr_locked+0x1d3/0x260 fs/xattr.c:277 vfs_setxattr+0x143/0x340 fs/xattr.c:309 setxattr+0x146/0x160 fs/xattr.c:617 path_setxattr+0x197/0x1c0 fs/xattr.c:636 __do_sys_setxattr fs/xattr.c:652 [inline] __se_sys_setxattr fs/xattr.c:648 [inline] __ia32_sys_setxattr+0xc0/0x160 fs/xattr.c:648 do_syscall_32_irqs_on arch/x86/entry/common.c:112 [inline] __do_fast_syscall_32+0x65/0xf0 arch/x86/entry/common.c:178 do_fast_syscall_32+0x33/0x70 arch/x86/entry/common.c:203 entry_SYSENTER_compat_after_hwframe+0x70/0x82Below is a similar scenario, the scenario in the syzbot log looks more complicated than this one, but this patch can fix it.
T21124
p9_read_work
======================== second trans =================================p9_client_walk
p9_client_rpcp9_client_prepare_req p9_tag_alloc req = kmem_cache_alloc(p9_req_cache, GFP_NOFS); tag = idr_alloc << preempted >> req->tc.tag = tag; /* req->[refcount/tag] == uninitialized */ m->rreq = p9_tag_lookup(m->client, m->rc.tag); /* increments uninitalized refcount */ refcount_set(&req->refcount, 2); /* cb drops one ref */ p9_client_cb(req) /* reader thread drops its ref: request is incorrectly freed */ p9_req_put(req) /* use after free and ref underflow */ p9_req_put(req)
Quoted 4 lines
To fix it, we can initize the refcount to zero before add to idr.
Reported-by [email protected]Signed-off-by Schspa Shi <[email protected]>
Quoted 17 lines
--
Changelog:
v1 -> v2:
- Set refcount to fix the problem.
v2 -> v3:
- Comment messages improve as asmadeus suggested.--- net/9p/client.c | 5 +++++ 1 file changed, 5 insertions(+)net/9p/client.c +0
-0
net/9p/client.c p9_tag_alloc(struct p9_client *c, int8_t type, uint t_size, uint r_size, · line 304
Quoted 8 lines
net/9p/client.c +5
-0@@ -297,6 +297,11 @@ p9_tag_alloc(struct p9_client *c, int8_t type, uint t_size, uint r_size,p9pdu_reset (&req->rc); req->t_err =0 ; req->status =REQ_STATUS_ALLOC ; +/* refcount needs to be set to 0 before inserting into the idr +* so p9_tag_lookup does not accept a request that is not fully +* initialized. refcount_set to 2 below will mark request live. +*/ +refcount_set (&req->refcount,0 );
I would s/live/ready for being used/, but comment should be clear enough anyway.
Quoted 4 lines
net/9p/client.c +0
-0init_waitqueue_head (&req->wq);INIT_LIST_HEAD (&req->req_list);
[email protected]
maintainerSchspa Shi wrote on Thu, Dec 01, 2022 at 11:33:10AM +0800:
Quoted 59 lines
When the new request allocated, the refcount will be zero if it is resued one. But if the request is newly allocated from slab, it is not fully initialized before add it to idr.
If the
p9_read_workgot a response before the refcount initiated. It will use a uninitialized req, which will result in a bad request data struct.Here is the logs from syzbot.
Corrupted memory at
0xffff88807eade00b[0xff0x070x000x000x000x000x000x00. . . . . . . . ] (in kfence-#110):p9_fcall_fini net/9p/client.c:248 [inline] p9_req_put net/9p/client.c:396 [inline] p9_req_put+0x208/0x250 net/9p/client.c:390 p9_client_walk+0x247/0x540 net/9p/client.c:1165 clone_fid fs/9p/fid.h:21 [inline] v9fs_fid_xattr_set+0xe4/0x2b0 fs/9p/xattr.c:118 v9fs_xattr_set fs/9p/xattr.c:100 [inline] v9fs_xattr_handler_set+0x6f/0x120 fs/9p/xattr.c:159 __vfs_setxattr+0x119/0x180 fs/xattr.c:182 __vfs_setxattr_noperm+0x129/0x5f0 fs/xattr.c:216 __vfs_setxattr_locked+0x1d3/0x260 fs/xattr.c:277 vfs_setxattr+0x143/0x340 fs/xattr.c:309 setxattr+0x146/0x160 fs/xattr.c:617 path_setxattr+0x197/0x1c0 fs/xattr.c:636 __do_sys_setxattr fs/xattr.c:652 [inline] __se_sys_setxattr fs/xattr.c:648 [inline] __ia32_sys_setxattr+0xc0/0x160 fs/xattr.c:648 do_syscall_32_irqs_on arch/x86/entry/common.c:112 [inline] __do_fast_syscall_32+0x65/0xf0 arch/x86/entry/common.c:178 do_fast_syscall_32+0x33/0x70 arch/x86/entry/common.c:203 entry_SYSENTER_compat_after_hwframe+0x70/0x82Below is a similar scenario, the scenario in the syzbot log looks more complicated than this one, but this patch can fix it.
T21124
p9_read_work
======================== second trans =================================p9_client_walk
p9_client_rpcp9_client_prepare_req p9_tag_alloc req = kmem_cache_alloc(p9_req_cache, GFP_NOFS); tag = idr_alloc << preempted >> req->tc.tag = tag; /* req->[refcount/tag] == uninitialized */ m->rreq = p9_tag_lookup(m->client, m->rc.tag); /* increments uninitalized refcount */ refcount_set(&req->refcount, 2); /* cb drops one ref */ p9_client_cb(req) /* reader thread drops its ref: request is incorrectly freed */ p9_req_put(req) /* use after free and ref underflow */ p9_req_put(req)
Quoted 1 line
To fix it, we can initize the refcount to zero before add to idr.
(fixed initialize typo here)
Quoted 10 lines
Reported-by [email protected]Signed-off-by Schspa Shi <[email protected]>--
Changelog:
v1 -> v2:
- Set refcount to fix the problem.
v2 -> v3:
- Comment messages improve as asmadeus suggested.
Just a note: when applying a patch with git am, this goes into the commit message -- please include the changelog below the git's three dashes instead (anything between the three dashes and the 'diff --git' below:
net/9p/client.c
Quoted 5 lines
--- net/9p/client.c | 5 +++++ 1 file changed, 5 insertions(+)net/9p/client.c +0
-0
Christian Schoenebeck wrote on Fri, Dec 02, 2022 at 12:48:39PM +0100:
Quoted 8 lines
net/9p/client.c p9_tag_alloc(struct p9_client *c, int8_t type, uint t_size, uint r_size, · line 304
Quoted 5 lines
net/9p/client.c +5
-0+/* refcount needs to be set to 0 before inserting into the idr +* so p9_tag_lookup does not accept a request that is not fully +* initialized. refcount_set to 2 below will mark request live. +*/ +refcount_set (&req->refcount,0 );I would s/live/ready for being used/, but comment should be clear enough anyway.
I blame golfing to fit into three lines, sorry!
Since it was my suggestion, I've taken the liberty to change 'live' to 'ready' as an half step; I think it's clearer than live and probably understandable enough.
I've pushed this to my next branch and will submit to Linus for the merge window in a couple of weeks, no point in rushing this to stable unless it gets snatched through the net tree first...
Signature · 1 line
Dominique
Quoted 75 lines · [email protected] writes:
Schspa Shi wrote on Thu, Dec 01, 2022 at 11:33:10AM +0800:
Quoted 59 lines
When the new request allocated, the refcount will be zero if it is resued one. But if the request is newly allocated from slab, it is not fully initialized before add it to idr.
If the
p9_read_workgot a response before the refcount initiated. It will use a uninitialized req, which will result in a bad request data struct.Here is the logs from syzbot.
Corrupted memory at
0xffff88807eade00b[0xff0x070x000x000x000x000x000x00. . . . . . . . ] (in kfence-#110):p9_fcall_fini net/9p/client.c:248 [inline] p9_req_put net/9p/client.c:396 [inline] p9_req_put+0x208/0x250 net/9p/client.c:390 p9_client_walk+0x247/0x540 net/9p/client.c:1165 clone_fid fs/9p/fid.h:21 [inline] v9fs_fid_xattr_set+0xe4/0x2b0 fs/9p/xattr.c:118 v9fs_xattr_set fs/9p/xattr.c:100 [inline] v9fs_xattr_handler_set+0x6f/0x120 fs/9p/xattr.c:159 __vfs_setxattr+0x119/0x180 fs/xattr.c:182 __vfs_setxattr_noperm+0x129/0x5f0 fs/xattr.c:216 __vfs_setxattr_locked+0x1d3/0x260 fs/xattr.c:277 vfs_setxattr+0x143/0x340 fs/xattr.c:309 setxattr+0x146/0x160 fs/xattr.c:617 path_setxattr+0x197/0x1c0 fs/xattr.c:636 __do_sys_setxattr fs/xattr.c:652 [inline] __se_sys_setxattr fs/xattr.c:648 [inline] __ia32_sys_setxattr+0xc0/0x160 fs/xattr.c:648 do_syscall_32_irqs_on arch/x86/entry/common.c:112 [inline] __do_fast_syscall_32+0x65/0xf0 arch/x86/entry/common.c:178 do_fast_syscall_32+0x33/0x70 arch/x86/entry/common.c:203 entry_SYSENTER_compat_after_hwframe+0x70/0x82Below is a similar scenario, the scenario in the syzbot log looks more complicated than this one, but this patch can fix it.
T21124
p9_read_work
======================== second trans =================================p9_client_walk
p9_client_rpcp9_client_prepare_req p9_tag_alloc req = kmem_cache_alloc(p9_req_cache, GFP_NOFS); tag = idr_alloc << preempted >> req->tc.tag = tag; /* req->[refcount/tag] == uninitialized */ m->rreq = p9_tag_lookup(m->client, m->rc.tag); /* increments uninitalized refcount */ refcount_set(&req->refcount, 2); /* cb drops one ref */ p9_client_cb(req) /* reader thread drops its ref: request is incorrectly freed */ p9_req_put(req) /* use after free and ref underflow */ p9_req_put(req)Quoted 1 line
To fix it, we can initize the refcount to zero before add to idr.
(fixed initialize typo here)
Quoted 10 lines
Reported-by [email protected]Signed-off-by Schspa Shi <[email protected]>--
Changelog:
v1 -> v2:
- Set refcount to fix the problem.
v2 -> v3:
- Comment messages improve as asmadeus suggested.
Quoted 4 lines
Just a note: when applying a patch with git am, this goes into the commit message -- please include the changelog below the git's three dashes instead (anything between the three dashes and the 'diff --git' below:
Thanks for the reminder, I will pay attention to this next time.
Quoted 17 lines
net/9p/client.c
Quoted 5 lines
--- net/9p/client.c | 5 +++++ 1 file changed, 5 insertions(+)net/9p/client.c +0
-0Christian Schoenebeck wrote on Fri, Dec 02, 2022 at 12:48:39PM +0100:
Quoted 8 lines
net/9p/client.c p9_tag_alloc(struct p9_client *c, int8_t type, uint t_size, uint r_size, · line 304
Quoted 5 lines
net/9p/client.c +5
-0+/* refcount needs to be set to 0 before inserting into the idr +* so p9_tag_lookup does not accept a request that is not fully +* initialized. refcount_set to 2 below will mark request live. +*/ +refcount_set (&req->refcount,0 );I would s/live/ready for being used/, but comment should be clear enough anyway.
Quoted 8 lines
I blame golfing to fit into three lines, sorry!
Since it was my suggestion, I've taken the liberty to change 'live' to 'ready' as an half step; I think it's clearer than live and probably understandable enough.I've pushed this to my next branch and will submit to Linus for the merge window in a couple of weeks, no point in rushing this to stable unless it gets snatched through the net tree first...
Thanks.
Signature · 2 lines
BRs Schspa Shi