PT-2026-151739
CVE-2026-99528
·
تاريخ النشر
2026-11-19
·
تاريخ التحديث
2026-11-19
الوصف من
الدرجة
N/A
غير متوفرلا تتوفر حاليًا تقييمات أو مقاييس للخطورة. سنحدّث المعلومات المقابلة في الصفحة فور توفرها.
In the Linux kernel, the following vulnerability has been resolved:
SUNRPC: fix netns use-after-free in write gssp()
While fuzzing with a customized syzkaller, I hit a refcount warning in
xprt alloc(), reached from a write to /proc/net/rpc/use-gss-proxy:
refcount t: addition on 0; use-after-free.
WARNING: lib/refcount.c:25 at refcount warn saturate+0xf8/0x120
CPU: 0 PID: 10404 Comm: syz.0.17 Not tainted 7.3.0-rc1-00096-gcfebfd3db73d
Call Trace:
xprt alloc+0x83f/0x9d0
xs setup xprt+0xaf/0x3c0
xs setup local+0x47/0x7f0
xprt create transport+0x16c/0x730
rpc create+0x38e/0x7f0
gssp rpc create+0xe2/0x180
set gssp clnt+0xba/0x1b0
write gssp+0x200/0x310
proc reg write+0x240/0x330
vfs write+0x2aa/0x1050
ksys write+0x12a/0x250
do syscall 64+0x117/0x750
entry SYSCALL 64 after hwframe+0x77/0x7f
create use gss proxy proc entry() stores the struct net pointer as the
proc entry's private data without taking a reference on it, and an open
file descriptor does not pin the namespace either, as procfs only pins
the proc dir entry. write gssp() reads that pointer back with
pde data() and hands it to set gssp clnt(), which reaches xprt init():
xprt->xprt net = get net track(net, &xprt->ns tracker, GFP KERNEL);get net track() -> get net() -> refcount inc() is unconditional, so
rpc create() assumes its caller holds a reference on args.net.
write gssp() does not.
Dropping the last ns.count reference does not make the file go away.
put net() only queues cleanup net() on a workqueue, and the
remove proc entry() that fences off further writes runs from the
rpcsec gss pernet .exit method, that is, from inside cleanup net(). A
write landing between those two points increments a refcount that is
already zero. cleanup net() does not re-read ns.count, so the namespace
is freed anyway and the rpc xprt -- along with the AF LOCAL socket
opened for it -- is left pointing at freed memory. The resurrected
count is visible to the rest of that teardown too: with
CONFIG IPV6 MROUTE the same cleanup net() worker then trips
!mr can free table() in ip6mr free table(), which tests check net() on
the namespace it is freeing.
It reduces to opening /proc/net/rpc/use-gss-proxy inside a new network
namespace, calling setns() back to the original one to drop the last
reference, and then writing "1" to the still-open descriptor.
Take the reference in write gssp() itself and refuse the write when the
namespace is already gone. This is what procfs does for every other
/proc/net file: get proc net() is maybe get net(PDE NET(PDE(inode)))
and seq open net() returns -ENXIO when it fails. PDE NET() cannot be
reused here because the parent directory /proc/net/rpc carries no
namespace pointer. Adding a .pre exit to rpcsec gss net ops so the
entry is removed earlier would only narrow the window, since pre exit
also runs from cleanup net().
Live namespaces are unaffected: a write still returns the error from
set gssp clnt(), a write of "2" still returns -EINVAL, and reads are
untouched.
وجدت خطأً في الوصف؟ لديك ما تضيفه؟ لا تتردد في مراسلتنا 👾