Message-ID: <CANT5p=qiuF0=PZaN9eCzrps=azamWjkh9U19zu2B5dRADUrP1g@mail.gmail.com>
Date: Thu, 7 Aug 2025 11:51:55 +0530
From: Shyam Prasad N <nspmangalore@...il.com>
To: Steve French <smfrench@...il.com>
Cc: Yunseong Kim <ysk@...lloc.com>, Steve French <sfrench@...ba.org>, 
	Paulo Alcantara <pc@...guebit.org>, Ronnie Sahlberg <ronniesahlberg@...il.com>, 
	Shyam Prasad N <sprasad@...rosoft.com>, Tom Talpey <tom@...pey.com>, 
	Bharath SM <bharathsm@...rosoft.com>, Namjae Jeon <linkinjeon@...nel.org>, 
	linux-cifs@...r.kernel.org, samba-technical@...ts.samba.org, 
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] cifs: Fix null-ptr-deref by static initializing global lock

On Thu, Aug 7, 2025 at 1:44 AM Steve French <smfrench@...il.com> wrote:
>
> merged into cifs-2.6.git for-next
>
> On Wed, Aug 6, 2025 at 8:23 AM Yunseong Kim <ysk@...lloc.com> wrote:
> >
> > A kernel panic can be triggered by reading /proc/fs/cifs/debug_dirs.
> > The crash is a null-ptr-deref inside spin_lock(), caused by the use of the
> > uninitialized global spinlock cifs_tcp_ses_lock.
> >
> > init_cifs()
> >  └── cifs_proc_init()
> >       └── // User can access /proc/fs/cifs/debug_dirs here
> >            └── cifs_debug_dirs_proc_show()
> >                 └── spin_lock(&cifs_tcp_ses_lock); // Uninitialized!
> >
> > KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007]
> > Mem abort info:
> > ESR = 0x0000000096000005
> > EC = 0x25: DABT (current EL), IL = 32 bits
> > SET = 0, FnV = 0
> > EA = 0, S1PTW = 0
> > FSC = 0x05: level 1 translation fault
> > Data abort info:
> > ISV = 0, ISS = 0x00000005, ISS2 = 0x00000000
> > CM = 0, WnR = 0, TnD = 0, TagAccess = 0
> > GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
> > [dfff800000000000] address between user and kernel address ranges
> > Internal error: Oops: 0000000096000005 [#1] SMP
> > Modules linked in:
> > CPU: 3 UID: 0 PID: 16435 Comm: stress-ng-procf Not tainted 6.16.0-10385-g79f14b5d84c6 #37 PREEMPT
> > Hardware name: QEMU KVM Virtual Machine, BIOS 2025.02-8ubuntu1 06/11/2025
> > pstate: 23400005 (nzCv daif +PAN -UAO +TCO +DIT -SSBS BTYPE=--)
> > pc : do_raw_spin_lock+0x84/0x2cc
> > lr : _raw_spin_lock+0x24/0x34
> > sp : ffff8000966477e0
> > x29: ffff800096647860 x28: ffff800096647b88 x27: ffff0001c0c22070
> > x26: ffff0003eb2b60c8 x25: ffff0001c0c22018 x24: dfff800000000000
> > x23: ffff0000f624e000 x22: ffff0003eb2b6020 x21: ffff0000f624e768
> > x20: 0000000000000004 x19: 0000000000000000 x18: 0000000000000000
> > x17: 0000000000000000 x16: ffff8000804b9600 x15: ffff700012cc8f04
> > x14: 1ffff00012cc8f04 x13: 0000000000000004 x12: ffffffffffffffff
> > x11: 1ffff00012cc8f00 x10: ffff80008d9af0d2 x9 : f3f3f304f1f1f1f1
> > x8 : 0000000000000000 x7 : 7365733c203e6469 x6 : 20656572743c2023
> > x5 : ffff0000e0ce0044 x4 : ffff80008a4deb6e x3 : ffff8000804b9718
> > x2 : 0000000000000001 x1 : 0000000000000000 x0 : 0000000000000000
> > Call trace:
> > do_raw_spin_lock+0x84/0x2cc (P)
> > _raw_spin_lock+0x24/0x34
> > cifs_debug_dirs_proc_show+0x1ac/0x4c0
> > seq_read_iter+0x3b0/0xc28
> > proc_reg_read_iter+0x178/0x2a8
> > vfs_read+0x5f8/0x88c
> > ksys_read+0x120/0x210
> > __arm64_sys_read+0x7c/0x90
> > invoke_syscall+0x98/0x2b8
> > el0_svc_common+0x130/0x23c
> > do_el0_svc+0x48/0x58
> > el0_svc+0x40/0x140
> > el0t_64_sync_handler+0x84/0x12c
> > el0t_64_sync+0x1ac/0x1b0
> > Code: aa0003f3 f9000feb f2fe7e69 f8386969 (38f86908)
> > ---[ end trace 0000000000000000 ]---
> >
> > The root cause is an initialization order problem. The lock is declared
> > as a global variable and intended to be initialized during module startup.
> > However, the procfs entry that uses this lock can be accessed by userspace
> > before the spin_lock_init() call has run. This creates a race window where
> > reading the proc file will attempt to use the lock before it is
> > initialized, leading to the crash.
> >
> > For a global lock with a static lifetime, the correct and robust approach
> > is to use compile-time initialization.
> >
> > Fixes: 844e5c0eb176 ("smb3 client: add way to show directory leases for improved debugging")
> > Signed-off-by: Yunseong Kim <ysk@...lloc.com>
> > ---
> >  fs/smb/client/cifsfs.c | 6 ++----
> >  1 file changed, 2 insertions(+), 4 deletions(-)
> >
> > diff --git a/fs/smb/client/cifsfs.c b/fs/smb/client/cifsfs.c
> > index 31930b7266db..3bd85ab2deb1 100644
> > --- a/fs/smb/client/cifsfs.c
> > +++ b/fs/smb/client/cifsfs.c
> > @@ -77,7 +77,7 @@ unsigned int global_secflags = CIFSSEC_DEF;
> >  unsigned int GlobalCurrentXid; /* protected by GlobalMid_Lock */
> >  unsigned int GlobalTotalActiveXid; /* prot by GlobalMid_Lock */
> >  unsigned int GlobalMaxActiveXid;       /* prot by GlobalMid_Lock */
> > -spinlock_t GlobalMid_Lock; /* protects above & list operations on midQ entries */
> > +DEFINE_SPINLOCK(GlobalMid_Lock); /* protects above & list operations on midQ entries */
> >
> >  /*
> >   *  Global counters, updated atomically
> > @@ -97,7 +97,7 @@ atomic_t total_buf_alloc_count;
> >  atomic_t total_small_buf_alloc_count;
> >  #endif/* STATS2 */
> >  struct list_head       cifs_tcp_ses_list;
> > -spinlock_t             cifs_tcp_ses_lock;
> > +DEFINE_SPINLOCK(cifs_tcp_ses_lock);
> >  static const struct super_operations cifs_super_ops;
> >  unsigned int CIFSMaxBufSize = CIFS_MAX_MSGSIZE;
> >  module_param(CIFSMaxBufSize, uint, 0444);
> > @@ -1863,8 +1863,6 @@ init_cifs(void)
> >         GlobalCurrentXid = 0;
> >         GlobalTotalActiveXid = 0;
> >         GlobalMaxActiveXid = 0;
> > -       spin_lock_init(&cifs_tcp_ses_lock);
> > -       spin_lock_init(&GlobalMid_Lock);
> >
> >         cifs_lock_secret = get_random_u32();
> >
> > --
> > 2.50.0
> >
> >
>
>
> --
> Thanks,
>
> Steve
>

Good catch.
But the problem is that cifs_proc_init gets called very early on in init_cifs.
That call should be moved to just before (or possibly after)
registering the filesystem.

-- 
Regards,
Shyam

Powered by blists - more mailing lists