Skip to contentSkip to conversation

Kernel oops: NULL pointer dereference in cifs_ioctl on 2.6.37-rc1

DISCUSSION Kjell Rune Skaaraas 4 participants 8 messages Nov 7, 2010 – Nov 8, 2010 6 messages not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

After upgrading from 2.6.36 for other reasons, starting certain apps like wine utorrent.exe will cause a kernel oops. I run the x86_64 version of Ubuntu 10.10 with various modified packages all around and the 2.6.37-rc1 kernel from the kernel PPA team. I experienced the same with a kernel I tried compiling myself too.

Nov  7 17:25:50 wodan kernel: [77498.450787] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040
Nov  7 17:25:50 wodan kernel: [77498.450883] IP: [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]
Nov  7 17:25:50 wodan kernel: [77498.450987] PGD 1836a067 PUD 5bfac067 PMD 17d2b067 PTE 0
Nov  7 17:25:50 wodan kernel: [77498.451234] CPU 4 
Nov  7 17:25:50 wodan kernel: [77498.451264] Modules linked in: binfmt_misc vboxnetadp vboxnetflt vboxdrv snd_hda_codec_hdmi snd_hda_codec_realtek nls_cp437 cifs snd_hda_intel snd_hda_codec snd_seq_midi snd_rawmidi snd_hwdep snd_pcm snd_seq_midi_event snd_seq snd_timer snd_seq_device ppdev lp parport_pc serio_raw snd shpchp soundcore parport joydev snd_page_alloc sha256_generic cryptd aes_x86_64 aes_generic dm_crypt radeon ttm usbhid ahci hid drm_kms_helper r8169 libahci pata_jmicron drm mii floppy i2c_algo_bit
Nov  7 17:25:50 wodan kernel: [77498.451845] 
Nov  7 17:25:50 wodan kernel: [77498.451865] Pid: 4386, comm: utorrent.exe Not tainted 2.6.37-020637rc1-generic #201011020905 P55-UD3R/P55-UD3R
Nov  7 17:25:50 wodan kernel: [77498.451971] RIP: 0010:[<ffffffffa0395729>]  [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]
Nov  7 17:25:50 wodan kernel: [77498.452074] RSP: 0018:ffff880018899e98  EFLAGS: 00010282
Nov  7 17:25:50 wodan kernel: [77498.452131] RAX: ffff880207a08300 RBX: ffff88005c884780 RCX: 0000000000222000
Nov  7 17:25:50 wodan kernel: [77498.452207] RDX: 0000000000222000 RSI: 0000000082187201 RDI: ffff88005c884780
Nov  7 17:25:50 wodan kernel: [77498.452282] RBP: ffff880018899ee8 R08: 000000000000001c R09: 00000000007bf700
Nov  7 17:25:50 wodan kernel: [77498.452358] R10: ffff880018898000 R11: 0000000000000000 R12: 0000000082187201
Nov  7 17:25:50 wodan kernel: [77498.452433] R13: 0000000000000000 R14: 0000000000222000 R15: ffff880207a4c2c0
Nov  7 17:25:50 wodan kernel: [77498.452510] FS:  000000007ffd8000(0063) GS:ffff8800df900000(006b) knlGS:00000000f74426c0
Nov  7 17:25:50 wodan kernel: [77498.452595] CS:  0010 DS: 002b ES: 002b CR0: 000000008005003b
Nov  7 17:25:50 wodan kernel: [77498.452656] CR2: 0000000000000040 CR3: 000000005b77e000 CR4: 00000000000026e0
Nov  7 17:25:50 wodan kernel: [77498.452732] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
Nov  7 17:25:50 wodan kernel: [77498.452808] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Nov  7 17:25:50 wodan kernel: [77498.452884] Process utorrent.exe (pid: 4386, threadinfo ffff880018898000, task ffff8800188bdb00)
Nov  7 17:25:50 wodan kernel: [77498.452999]  00000000fffffff7 ffff880018899ef8 0000000000000000 0000000000000000
Nov  7 17:25:50 wodan kernel: [77498.453084]  ffff880018899ee8 0000000000000000 0000000000222000 ffff88005c884780
Nov  7 17:25:50 wodan kernel: [77498.453169]  0000000000222000 000000000000001c ffff880018899ef8 ffffffff8116e94d
Nov  7 17:25:50 wodan kernel: [77498.453287]  [<ffffffff8116e94d>] vfs_ioctl+0x1d/0x50
Nov  7 17:25:50 wodan kernel: [77498.453343]  [<ffffffff8116f52d>] do_vfs_ioctl+0x7d/0x1f0
Nov  7 17:25:50 wodan kernel: [77498.453405]  [<ffffffff811a5c62>] compat_sys_ioctl+0x122/0x200
Nov  7 17:25:50 wodan kernel: [77498.453470]  [<ffffffff81048560>] sysenter_dispatch+0x7/0x2e
Nov  7 17:25:50 wodan kernel: [77498.454010]  RSP <ffff880018899e98>
Nov  7 17:25:50 wodan kernel: [77498.478707] ---[ end trace 4e689dcb8ba02a93 ]---

Hopefully this makes more sense to someone than it does to me :)

Regards,
Kjell Rune

A message that is not in the archive (sent off-list or not yet received): 484246.91210.qm-ZxlQ8pIuIvXGRxTy+Q50vsz6deESKz/[email protected]

#2 lore raw
Quoted 32 lines · On Sun, 7 Nov 2010 16:44:46 +0000 (GMT) Kjell Rune Skaaraas…

After upgrading from 2.6.36 for other reasons, starting certain apps like wine utorrent.exe will cause a kernel oops. I run the x86_64 version of Ubuntu 10.10 with various modified packages all around and the 2.6.37-rc1 kernel from the kernel PPA team. I experienced the same with a kernel I tried compiling myself too.

Nov  7 17:25:50 wodan kernel: [77498.450787] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040
Nov  7 17:25:50 wodan kernel: [77498.450883] IP: [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]
Nov  7 17:25:50 wodan kernel: [77498.450987] PGD 1836a067 PUD 5bfac067 PMD 17d2b067 PTE 0
Nov  7 17:25:50 wodan kernel: [77498.451234] CPU 4 
Nov  7 17:25:50 wodan kernel: [77498.451264] Modules linked in: binfmt_misc vboxnetadp vboxnetflt vboxdrv snd_hda_codec_hdmi snd_hda_codec_realtek nls_cp437 cifs snd_hda_intel snd_hda_codec snd_seq_midi snd_rawmidi snd_hwdep snd_pcm snd_seq_midi_event snd_seq snd_timer snd_seq_device ppdev lp parport_pc serio_raw snd shpchp soundcore parport joydev snd_page_alloc sha256_generic cryptd aes_x86_64 aes_generic dm_crypt radeon ttm usbhid ahci hid drm_kms_helper r8169 libahci pata_jmicron drm mii floppy i2c_algo_bit
Nov  7 17:25:50 wodan kernel: [77498.451845] 
Nov  7 17:25:50 wodan kernel: [77498.451865] Pid: 4386, comm: utorrent.exe Not tainted 2.6.37-020637rc1-generic #201011020905 P55-UD3R/P55-UD3R
Nov  7 17:25:50 wodan kernel: [77498.451971] RIP: 0010:[<ffffffffa0395729>]  [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]
Nov  7 17:25:50 wodan kernel: [77498.452074] RSP: 0018:ffff880018899e98  EFLAGS: 00010282
Nov  7 17:25:50 wodan kernel: [77498.452131] RAX: ffff880207a08300 RBX: ffff88005c884780 RCX: 0000000000222000
Nov  7 17:25:50 wodan kernel: [77498.452207] RDX: 0000000000222000 RSI: 0000000082187201 RDI: ffff88005c884780
Nov  7 17:25:50 wodan kernel: [77498.452282] RBP: ffff880018899ee8 R08: 000000000000001c R09: 00000000007bf700
Nov  7 17:25:50 wodan kernel: [77498.452358] R10: ffff880018898000 R11: 0000000000000000 R12: 0000000082187201
Nov  7 17:25:50 wodan kernel: [77498.452433] R13: 0000000000000000 R14: 0000000000222000 R15: ffff880207a4c2c0
Nov  7 17:25:50 wodan kernel: [77498.452510] FS:  000000007ffd8000(0063) GS:ffff8800df900000(006b) knlGS:00000000f74426c0
Nov  7 17:25:50 wodan kernel: [77498.452595] CS:  0010 DS: 002b ES: 002b CR0: 000000008005003b
Nov  7 17:25:50 wodan kernel: [77498.452656] CR2: 0000000000000040 CR3: 000000005b77e000 CR4: 00000000000026e0
Nov  7 17:25:50 wodan kernel: [77498.452732] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
Nov  7 17:25:50 wodan kernel: [77498.452808] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Nov  7 17:25:50 wodan kernel: [77498.452884] Process utorrent.exe (pid: 4386, threadinfo ffff880018898000, task ffff8800188bdb00)
Nov  7 17:25:50 wodan kernel: [77498.452999]  00000000fffffff7 ffff880018899ef8 0000000000000000 0000000000000000
Nov  7 17:25:50 wodan kernel: [77498.453084]  ffff880018899ee8 0000000000000000 0000000000222000 ffff88005c884780
Nov  7 17:25:50 wodan kernel: [77498.453169]  0000000000222000 000000000000001c ffff880018899ef8 ffffffff8116e94d
Nov  7 17:25:50 wodan kernel: [77498.453287]  [<ffffffff8116e94d>] vfs_ioctl+0x1d/0x50
Nov  7 17:25:50 wodan kernel: [77498.453343]  [<ffffffff8116f52d>] do_vfs_ioctl+0x7d/0x1f0
Nov  7 17:25:50 wodan kernel: [77498.453405]  [<ffffffff811a5c62>] compat_sys_ioctl+0x122/0x200
Nov  7 17:25:50 wodan kernel: [77498.453470]  [<ffffffff81048560>] sysenter_dispatch+0x7/0x2e
Nov  7 17:25:50 wodan kernel: [77498.454010]  RSP <ffff880018899e98>
Nov  7 17:25:50 wodan kernel: [77498.478707] ---[ end trace 4e689dcb8ba02a93 ]---
Quoted 2 lines

Hopefully this makes more sense to someone than it does to me :)

Could you follow the directions here and send along the output? That will tell us where it's panicing:

http://wiki.samba.org/index.php/LinuxCIFS_troubleshooting#Oopses

Many Thanks,

Signature · 1 line
Jeff Layton <[email protected]>

A message that is not in the archive (sent off-list or not yet received): 20101107211202.3b3468dd-4QP7MXygkU+dMjc06nkz3ljfA9RmPOcC@public.gmane.org

#4 lore raw

3 Changesets affected this function in 2.6.37, but nothing particularly obvious. From a quick glance at objdump it may be around the line:

cifs_sb = CIFS_SB(inode->i_sb);

Quoted 51 lines · Jeff Layton
Quoted 32 lines · On Sun, 7 Nov 2010 16:44:46 +0000 (GMT) Kjell Rune Skaaraas…

After upgrading from 2.6.36 for other reasons, starting certain apps like wine utorrent.exe will cause a kernel oops. I run the x86_64 version of Ubuntu 10.10 with various modified packages all around and the 2.6.37-rc1 kernel from the kernel PPA team. I experienced the same with a kernel I tried compiling myself too.

Nov  7 17:25:50 wodan kernel: [77498.450787] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040
Nov  7 17:25:50 wodan kernel: [77498.450883] IP: [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]
Nov  7 17:25:50 wodan kernel: [77498.450987] PGD 1836a067 PUD 5bfac067 PMD 17d2b067 PTE 0
Nov  7 17:25:50 wodan kernel: [77498.451234] CPU 4
Nov  7 17:25:50 wodan kernel: [77498.451264] Modules linked in: binfmt_misc vboxnetadp vboxnetflt vboxdrv snd_hda_codec_hdmi snd_hda_codec_realtek nls_cp437 cifs snd_hda_intel snd_hda_codec snd_seq_midi snd_rawmidi snd_hwdep snd_pcm snd_seq_midi_event snd_seq snd_timer snd_seq_device ppdev lp parport_pc serio_raw snd shpchp soundcore parport joydev snd_page_alloc sha256_generic cryptd aes_x86_64 aes_generic dm_crypt radeon ttm usbhid ahci hid drm_kms_helper r8169 libahci pata_jmicron drm mii floppy i2c_algo_bit Nov  7 17:25:50 wodan kernel: [77498.451845]
Nov  7 17:25:50 wodan kernel: [77498.451865] Pid: 4386, comm: utorrent.exe Not tainted 2.6.37-020637rc1-generic #201011020905 P55-UD3R/P55-UD3R
Nov  7 17:25:50 wodan kernel: [77498.451971] RIP: 0010:[<ffffffffa0395729>]  [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]
Nov  7 17:25:50 wodan kernel: [77498.452074] RSP: 0018:ffff880018899e98  EFLAGS: 00010282
Nov  7 17:25:50 wodan kernel: [77498.452131] RAX: ffff880207a08300 RBX: ffff88005c884780 RCX: 0000000000222000
Nov  7 17:25:50 wodan kernel: [77498.452207] RDX: 0000000000222000 RSI: 0000000082187201 RDI: ffff88005c884780
Nov  7 17:25:50 wodan kernel: [77498.452282] RBP: ffff880018899ee8 R08: 000000000000001c R09: 00000000007bf700
Nov  7 17:25:50 wodan kernel: [77498.452358] R10: ffff880018898000 R11: 0000000000000000 R12: 0000000082187201
Nov  7 17:25:50 wodan kernel: [77498.452433] R13: 0000000000000000 R14: 0000000000222000 R15: ffff880207a4c2c0
Nov  7 17:25:50 wodan kernel: [77498.452510] FS:  000000007ffd8000(0063) GS:ffff8800df900000(006b) knlGS:00000000f74426c0
Nov  7 17:25:50 wodan kernel: [77498.452595] CS:  0010 DS: 002b ES: 002b CR0: 000000008005003b
Nov  7 17:25:50 wodan kernel: [77498.452656] CR2: 0000000000000040 CR3: 000000005b77e000 CR4: 00000000000026e0
Nov  7 17:25:50 wodan kernel: [77498.452732] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
Nov  7 17:25:50 wodan kernel: [77498.452808] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Nov  7 17:25:50 wodan kernel: [77498.452884] Process utorrent.exe (pid: 4386, threadinfo ffff880018898000, task ffff8800188bdb00)
Nov  7 17:25:50 wodan kernel: [77498.452999]  00000000fffffff7 ffff880018899ef8 0000000000000000 0000000000000000
Nov  7 17:25:50 wodan kernel: [77498.453084]  ffff880018899ee8 0000000000000000 0000000000222000 ffff88005c884780
Nov  7 17:25:50 wodan kernel: [77498.453169]  0000000000222000 000000000000001c ffff880018899ef8 ffffffff8116e94d
Nov  7 17:25:50 wodan kernel: [77498.453287]  [<ffffffff8116e94d>] vfs_ioctl+0x1d/0x50
Nov  7 17:25:50 wodan kernel: [77498.453343]  [<ffffffff8116f52d>] do_vfs_ioctl+0x7d/0x1f0
Nov  7 17:25:50 wodan kernel: [77498.453405]  [<ffffffff811a5c62>] compat_sys_ioctl+0x122/0x200
Nov  7 17:25:50 wodan kernel: [77498.453470]  [<ffffffff81048560>] sysenter_dispatch+0x7/0x2e
Nov  7 17:25:50 wodan kernel: [77498.454010]  RSP <ffff880018899e98>
Nov  7 17:25:50 wodan kernel: [77498.478707] ---[ end trace 4e689dcb8ba02a93 ]---

Quoted 2 lines

Hopefully this makes more sense to someone than it does to me :)

Could you follow the directions here and send along the output? That will tell us where it's panicing:

http://wiki.samba.org/index.php/LinuxCIFS_troubleshooting#Oopses

Many Thanks,
--
Jeff Layton <[email protected]> --
To unsubscribe from this list: send the line "unsubscribe linux-cifs" in the body of a message to [email protected] More majordomo info at  http://vger.kernel.org/majordomo-info.html

Signature · 3 lines
Thanks,

Steve

A message that is not in the archive (sent off-list or not yet received): AANLkTikzTztJUP5-zHCKr4nzuG=bSzmZ+ga7dU0r3QPU-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#6 lore raw
Quoted 8 lines · Steve French
Quoted 7 lines · Jeff Layton
Quoted 4 lines · On Sun, 7 Nov 2010 16:44:46 +0000 (GMT) Kjell Rune Skaaraas…

After upgrading from 2.6.36 for other reasons, starting certain apps like wine utorrent.exe will cause a kernel oops. I run the x86_64 version of Ubuntu 10.10 with various modified packages all around and the 2.6.37-rc1 kernel from the kernel PPA team. I experienced the same with a kernel I tried compiling myself too.

Nov ý7 17:25:50 wodan kernel: [77498.450787] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040 Nov ý7 17:25:50 wodan kernel: [77498.450883] IP: [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]

Does the below patch fixes your problem?

From: Suresh Jayaraman <[email protected]>
Subject: [PATCH] cifs: fix a NULL pointer dereference in cifs_ioctl() when the fd is bad

The commit ba00ba modified cifs_ioctl() to use tcon pointer in cifsFileInfo via tlink instead of cifs_sb->tcon. When the file handle is not valid the cifsFileInfo->tlink will be NULL. Fix this by getting the tcon pointer by calling cifs_sb_master_tcon().

Here's a hackish reproducer:

#include  <fcntl.h>
#include  <sys/ioctl.h>
#include  <sys/stat.h>
#include  <sys/types.h>
#include  <unistd.h>

#define CIFS_IOC_CHECKUMOUNT _IO(0xCF, 2)

int main (int argc, char* argv[])

 {
    int fd = open (argv[1], O_RDWR);

    ioctl(fd, CIFS_IOC_CHECKUMOUNT);

    close(fd);
    return 0;
}

This program will cause an oops when called with cifs mount point as an argument. I have tested the fix with the reproducer and it no longer oopses.

---
 fs/cifs/ioctl.c |    6 ++----
 1 files changed, 2 insertions(+), 4 deletions(-)
fs/cifs/ioctl.c +2 -4
index 2fa22f2..b8f680a 100644
@@ -35,10 +35,10 @@ long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg)
 	struct inode *inode = filep->f_dentry->d_inode;
 	int rc = -ENOTTY; /* strange error - but the precedent */
 	int xid;
-	struct cifs_sb_info *cifs_sb;
+	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
 #ifdef CONFIG_CIFS_POSIX
 	struct cifsFileInfo *pSMBFile = filep->private_data;
-	struct cifsTconInfo *tcon = tlink_tcon(pSMBFile->tlink);
+	struct cifsTconInfo *tcon = cifs_sb_master_tcon(cifs_sb);
 	__u64	ExtAttrBits = 0;
 	__u64	ExtAttrMask = 0;
 	__u64   caps = le64_to_cpu(tcon->fsUnixInfo.Capability);
@@ -48,8 +48,6 @@ long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg)
 
 	cFYI(1, "ioctl file %p  cmd %u  arg %lu", filep, command, arg);
 
-	cifs_sb = CIFS_SB(inode->i_sb);
-
 	switch (command) {
 		case CIFS_IOC_CHECKUMOUNT:
 			cFYI(1, "User unmount attempted");

A message that is not in the archive (sent off-list or not yet received): [email protected]

#8 lore raw
Quoted 70 lines · On Mon, 08 Nov 2010 14:05:03 +0530 Suresh Jayaraman…
Quoted 8 lines · Steve French
Quoted 7 lines · Jeff Layton
Quoted 4 lines · On Sun, 7 Nov 2010 16:44:46 +0000 (GMT) Kjell Rune Skaaraas…

After upgrading from 2.6.36 for other reasons, starting certain apps like wine utorrent.exe will cause a kernel oops. I run the x86_64 version of Ubuntu 10.10 with various modified packages all around and the 2.6.37-rc1 kernel from the kernel PPA team. I experienced the same with a kernel I tried compiling myself too.

Nov ý7 17:25:50 wodan kernel: [77498.450787] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040 Nov ý7 17:25:50 wodan kernel: [77498.450883] IP: [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]

Does the below patch fixes your problem?

From: Suresh Jayaraman <[email protected]>
Subject: [PATCH] cifs: fix a NULL pointer dereference in cifs_ioctl() when the fd is bad

The commit ba00ba modified cifs_ioctl() to use tcon pointer in cifsFileInfo via tlink instead of cifs_sb->tcon. When the file handle is not valid the cifsFileInfo->tlink will be NULL. Fix this by getting the tcon pointer by calling cifs_sb_master_tcon().

Here's a hackish reproducer:

#include  <fcntl.h>
#include  <sys/ioctl.h>
#include  <sys/stat.h>
#include  <sys/types.h>
#include  <unistd.h>

#define CIFS_IOC_CHECKUMOUNT _IO(0xCF, 2)

int main (int argc, char* argv[])

 {
    int fd = open (argv[1], O_RDWR);

    ioctl(fd, CIFS_IOC_CHECKUMOUNT);

    close(fd);
    return 0;
}

This program will cause an oops when called with cifs mount point as an argument. I have tested the fix with the reproducer and it no longer oopses.

---
 fs/cifs/ioctl.c |    6 ++----
 1 files changed, 2 insertions(+), 4 deletions(-)
fs/cifs/ioctl.c +2 -2
index 2fa22f2..b8f680a 100644
@@ -35,10 +35,10 @@ long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg)
 	struct inode *inode = filep->f_dentry->d_inode;
 	int rc = -ENOTTY; /* strange error - but the precedent */
 	int xid;
-	struct cifs_sb_info *cifs_sb;
+	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
 #ifdef CONFIG_CIFS_POSIX
 	struct cifsFileInfo *pSMBFile = filep->private_data;
-	struct cifsTconInfo *tcon = tlink_tcon(pSMBFile->tlink);
+	struct cifsTconInfo *tcon = cifs_sb_master_tcon(cifs_sb);
 	__u64	ExtAttrBits = 0;
 	__u64	ExtAttrMask = 0;
 	__u64   caps = le64_to_cpu(tcon->fsUnixInfo.Capability);

fs/cifs/ioctl.c long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg) · line 53

Quoted 8 lines
fs/cifs/ioctl.c +0 -2
@@ -48,8 +48,6 @@ long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg)
 
 	cFYI(1, "ioctl file %p  cmd %u  arg %lu", filep, command, arg);
 
-	cifs_sb = CIFS_SB(inode->i_sb);
-
 	switch (command) {
 		case CIFS_IOC_CHECKUMOUNT:
 			cFYI(1, "User unmount attempted");

NAK. This will mean that you're allowing people to do ioctls against files using other people's credentials. If this is the problem then the correct fix is to make sure that a cifsFileInfo is allocated during the open.

Signature · 1 line
Jeff Layton <[email protected]>

A message that is not in the archive (sent off-list or not yet received): 20101108061227.051706ff-9yPaYZwiELC+kQycOl6kW4xkIHaj4LzF@public.gmane.org

#10 lore raw
Quoted 76 lines · Jeff Layton
Quoted 73 lines · On Mon, 08 Nov 2010 14:05:03 +0530 Suresh Jayaraman…
Quoted 8 lines · Steve French
Quoted 7 lines · Jeff Layton
Quoted 4 lines · On Sun, 7 Nov 2010 16:44:46 +0000 (GMT) Kjell Rune Skaaraas…

After upgrading from 2.6.36 for other reasons, starting certain apps like wine utorrent.exe will cause a kernel oops. I run the x86_64 version of Ubuntu 10.10 with various modified packages all around and the 2.6.37-rc1 kernel from the kernel PPA team. I experienced the same with a kernel I tried compiling myself too.

Nov �7 17:25:50 wodan kernel: [77498.450787] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040 Nov �7 17:25:50 wodan kernel: [77498.450883] IP: [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]

Does the below patch fixes your problem?

From: Suresh Jayaraman <[email protected]>
Subject: [PATCH] cifs: fix a NULL pointer dereference in cifs_ioctl() when the fd is bad

The commit ba00ba modified cifs_ioctl() to use tcon pointer in cifsFileInfo via tlink instead of cifs_sb->tcon. When the file handle is not valid the cifsFileInfo->tlink will be NULL. Fix this by getting the tcon pointer by calling cifs_sb_master_tcon().

Here's a hackish reproducer:

#include  <fcntl.h>
#include  <sys/ioctl.h>
#include  <sys/stat.h>
#include  <sys/types.h>
#include  <unistd.h>

#define CIFS_IOC_CHECKUMOUNT _IO(0xCF, 2)

int main (int argc, char* argv[])

 {
    int fd = open (argv[1], O_RDWR);

    ioctl(fd, CIFS_IOC_CHECKUMOUNT);

    close(fd);
    return 0;
}

This program will cause an oops when called with cifs mount point as an argument. I have tested the fix with the reproducer and it no longer oopses.

---
 fs/cifs/ioctl.c |    6 ++----
 1 files changed, 2 insertions(+), 4 deletions(-)
fs/cifs/ioctl.c +2 -2
index 2fa22f2..b8f680a 100644
@@ -35,10 +35,10 @@ long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg)
 	struct inode *inode = filep->f_dentry->d_inode;
 	int rc = -ENOTTY; /* strange error - but the precedent */
 	int xid;
-	struct cifs_sb_info *cifs_sb;
+	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
 #ifdef CONFIG_CIFS_POSIX
 	struct cifsFileInfo *pSMBFile = filep->private_data;
-	struct cifsTconInfo *tcon = tlink_tcon(pSMBFile->tlink);
+	struct cifsTconInfo *tcon = cifs_sb_master_tcon(cifs_sb);
 	__u64	ExtAttrBits = 0;
 	__u64	ExtAttrMask = 0;
 	__u64   caps = le64_to_cpu(tcon->fsUnixInfo.Capability);
@@ -48,8 +48,6 @@ long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg)
 
 	cFYI(1, "ioctl file %p  cmd %u  arg %lu", filep, command, arg);
 
Quoted 7 lines

fs/cifs/ioctl.c long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg) · line 53

Quoted 5 lines
fs/cifs/ioctl.c +0 -2
-	cifs_sb = CIFS_SB(inode->i_sb);
-
 	switch (command) {
 		case CIFS_IOC_CHECKUMOUNT:
 			cFYI(1, "User unmount attempted");

NAK. This will mean that you're allowing people to do ioctls against

I missed this.. though I was not quite sure about this patch..

Quoted 1 line

files using other people's credentials. If this is the problem then the

I'm sure the problem is pSMBFile->tlink is NULL. And this is because open() fails and the test program didn't check for the error and continues using the return value (in this case -1 which will be UNSIGNED_INT_MAX) in the ioctl. Since the file is actually not open, cifs_new_fileinfo() won't have a chance to setup pSMBFile->tlink properly.

Signature · 1 line
Suresh Jayaraman

A message that is not in the archive (sent off-list or not yet received): [email protected]

#12 lore raw
Quoted 87 lines · On Mon, 08 Nov 2010 16:51:02 +0530 Suresh Jayaraman…
Quoted 76 lines · Jeff Layton
Quoted 73 lines · On Mon, 08 Nov 2010 14:05:03 +0530 Suresh Jayaraman…
Quoted 8 lines · Steve French
Quoted 7 lines · Jeff Layton
Quoted 4 lines · On Sun, 7 Nov 2010 16:44:46 +0000 (GMT) Kjell Rune Skaaraas…

After upgrading from 2.6.36 for other reasons, starting certain apps like wine utorrent.exe will cause a kernel oops. I run the x86_64 version of Ubuntu 10.10 with various modified packages all around and the 2.6.37-rc1 kernel from the kernel PPA team. I experienced the same with a kernel I tried compiling myself too.

Nov �7 17:25:50 wodan kernel: [77498.450787] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040 Nov �7 17:25:50 wodan kernel: [77498.450883] IP: [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]

Does the below patch fixes your problem?

From: Suresh Jayaraman <[email protected]>
Subject: [PATCH] cifs: fix a NULL pointer dereference in cifs_ioctl() when the fd is bad

The commit ba00ba modified cifs_ioctl() to use tcon pointer in cifsFileInfo via tlink instead of cifs_sb->tcon. When the file handle is not valid the cifsFileInfo->tlink will be NULL. Fix this by getting the tcon pointer by calling cifs_sb_master_tcon().

Here's a hackish reproducer:

#include  <fcntl.h>
#include  <sys/ioctl.h>
#include  <sys/stat.h>
#include  <sys/types.h>
#include  <unistd.h>

#define CIFS_IOC_CHECKUMOUNT _IO(0xCF, 2)

int main (int argc, char* argv[])

 {
    int fd = open (argv[1], O_RDWR);

    ioctl(fd, CIFS_IOC_CHECKUMOUNT);

    close(fd);
    return 0;
}

This program will cause an oops when called with cifs mount point as an argument. I have tested the fix with the reproducer and it no longer oopses.

---
 fs/cifs/ioctl.c |    6 ++----
 1 files changed, 2 insertions(+), 4 deletions(-)
fs/cifs/ioctl.c +2 -2
index 2fa22f2..b8f680a 100644
@@ -35,10 +35,10 @@ long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg)
 	struct inode *inode = filep->f_dentry->d_inode;
 	int rc = -ENOTTY; /* strange error - but the precedent */
 	int xid;
-	struct cifs_sb_info *cifs_sb;
+	struct cifs_sb_info *cifs_sb = CIFS_SB(inode->i_sb);
 #ifdef CONFIG_CIFS_POSIX
 	struct cifsFileInfo *pSMBFile = filep->private_data;
-	struct cifsTconInfo *tcon = tlink_tcon(pSMBFile->tlink);
+	struct cifsTconInfo *tcon = cifs_sb_master_tcon(cifs_sb);
 	__u64	ExtAttrBits = 0;
 	__u64	ExtAttrMask = 0;
 	__u64   caps = le64_to_cpu(tcon->fsUnixInfo.Capability);
@@ -48,8 +48,6 @@ long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg)
 
 	cFYI(1, "ioctl file %p  cmd %u  arg %lu", filep, command, arg);
 
Quoted 7 lines

fs/cifs/ioctl.c long cifs_ioctl(struct file *filep, unsigned int command, unsigned long arg) · line 53

Quoted 5 lines
fs/cifs/ioctl.c +0 -2
-	cifs_sb = CIFS_SB(inode->i_sb);
-
 	switch (command) {
 		case CIFS_IOC_CHECKUMOUNT:
 			cFYI(1, "User unmount attempted");

NAK. This will mean that you're allowing people to do ioctls against

I missed this.. though I was not quite sure about this patch..

Quoted 8 lines
Quoted 1 line

files using other people's credentials. If this is the problem then the

I'm sure the problem is pSMBFile->tlink is NULL. And this is because open() fails and the test program didn't check for the error and continues using the return value (in this case -1 which will be UNSIGNED_INT_MAX) in the ioctl. Since the file is actually not open, cifs_new_fileinfo() won't have a chance to setup pSMBFile->tlink properly.

That doesn't sound right. The VFS shouldn't allow an ioctl on an invalid file descriptor...

It seems more likely that pSMBFile is actually NULL here. cifs doesn't have f_op->open routine for directories, so that's entirely handled at the VFS layer. The right fix is probably to add one, and make sure that filp->private_data is appropriately populated on the open call for directories.

Signature · 1 line
Jeff Layton <[email protected]>
#13 lore raw
In reply to a message not in the archive
Quoted 32 lines · On Sun, 7 Nov 2010 16:44:46 +0000 (GMT) Kjell Rune Skaaraas…

After upgrading from 2.6.36 for other reasons, starting certain apps like wine utorrent.exe will cause a kernel oops. I run the x86_64 version of Ubuntu 10.10 with various modified packages all around and the 2.6.37-rc1 kernel from the kernel PPA team. I experienced the same with a kernel I tried compiling myself too.

Nov  7 17:25:50 wodan kernel: [77498.450787] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040
Nov  7 17:25:50 wodan kernel: [77498.450883] IP: [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]
Nov  7 17:25:50 wodan kernel: [77498.450987] PGD 1836a067 PUD 5bfac067 PMD 17d2b067 PTE 0
Nov  7 17:25:50 wodan kernel: [77498.451234] CPU 4 
Nov  7 17:25:50 wodan kernel: [77498.451264] Modules linked in: binfmt_misc vboxnetadp vboxnetflt vboxdrv snd_hda_codec_hdmi snd_hda_codec_realtek nls_cp437 cifs snd_hda_intel snd_hda_codec snd_seq_midi snd_rawmidi snd_hwdep snd_pcm snd_seq_midi_event snd_seq snd_timer snd_seq_device ppdev lp parport_pc serio_raw snd shpchp soundcore parport joydev snd_page_alloc sha256_generic cryptd aes_x86_64 aes_generic dm_crypt radeon ttm usbhid ahci hid drm_kms_helper r8169 libahci pata_jmicron drm mii floppy i2c_algo_bit
Nov  7 17:25:50 wodan kernel: [77498.451845] 
Nov  7 17:25:50 wodan kernel: [77498.451865] Pid: 4386, comm: utorrent.exe Not tainted 2.6.37-020637rc1-generic #201011020905 P55-UD3R/P55-UD3R
Nov  7 17:25:50 wodan kernel: [77498.451971] RIP: 0010:[<ffffffffa0395729>]  [<ffffffffa0395729>] cifs_ioctl+0x39/0x2f0 [cifs]
Nov  7 17:25:50 wodan kernel: [77498.452074] RSP: 0018:ffff880018899e98  EFLAGS: 00010282
Nov  7 17:25:50 wodan kernel: [77498.452131] RAX: ffff880207a08300 RBX: ffff88005c884780 RCX: 0000000000222000
Nov  7 17:25:50 wodan kernel: [77498.452207] RDX: 0000000000222000 RSI: 0000000082187201 RDI: ffff88005c884780
Nov  7 17:25:50 wodan kernel: [77498.452282] RBP: ffff880018899ee8 R08: 000000000000001c R09: 00000000007bf700
Nov  7 17:25:50 wodan kernel: [77498.452358] R10: ffff880018898000 R11: 0000000000000000 R12: 0000000082187201
Nov  7 17:25:50 wodan kernel: [77498.452433] R13: 0000000000000000 R14: 0000000000222000 R15: ffff880207a4c2c0
Nov  7 17:25:50 wodan kernel: [77498.452510] FS:  000000007ffd8000(0063) GS:ffff8800df900000(006b) knlGS:00000000f74426c0
Nov  7 17:25:50 wodan kernel: [77498.452595] CS:  0010 DS: 002b ES: 002b CR0: 000000008005003b
Nov  7 17:25:50 wodan kernel: [77498.452656] CR2: 0000000000000040 CR3: 000000005b77e000 CR4: 00000000000026e0
Nov  7 17:25:50 wodan kernel: [77498.452732] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
Nov  7 17:25:50 wodan kernel: [77498.452808] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Nov  7 17:25:50 wodan kernel: [77498.452884] Process utorrent.exe (pid: 4386, threadinfo ffff880018898000, task ffff8800188bdb00)
Nov  7 17:25:50 wodan kernel: [77498.452999]  00000000fffffff7 ffff880018899ef8 0000000000000000 0000000000000000
Nov  7 17:25:50 wodan kernel: [77498.453084]  ffff880018899ee8 0000000000000000 0000000000222000 ffff88005c884780
Nov  7 17:25:50 wodan kernel: [77498.453169]  0000000000222000 000000000000001c ffff880018899ef8 ffffffff8116e94d
Nov  7 17:25:50 wodan kernel: [77498.453287]  [<ffffffff8116e94d>] vfs_ioctl+0x1d/0x50
Nov  7 17:25:50 wodan kernel: [77498.453343]  [<ffffffff8116f52d>] do_vfs_ioctl+0x7d/0x1f0
Nov  7 17:25:50 wodan kernel: [77498.453405]  [<ffffffff811a5c62>] compat_sys_ioctl+0x122/0x200
Nov  7 17:25:50 wodan kernel: [77498.453470]  [<ffffffff81048560>] sysenter_dispatch+0x7/0x2e
Nov  7 17:25:50 wodan kernel: [77498.454010]  RSP <ffff880018899e98>
Nov  7 17:25:50 wodan kernel: [77498.478707] ---[ end trace 4e689dcb8ba02a93 ]---
Quoted 2 lines

Hopefully this makes more sense to someone than it does to me :)

Ok, the patch I just sent to the list seems to fix this for me and this is clearly our bug. But, I have to wonder -- what is wine doing running ioctls against files?

Signature · 1 line
Jeff Layton <[email protected]>