• S
    st: null pointer dereference panic caused by use after kref_put by st_open · e7ac6c66
    Seymour, Shane M 提交于
    Two SLES11 SP3 servers encountered similar crashes simultaneously
    following some kind of SAN/tape target issue:
    
    ...
    qla2xxx [0000:81:00.0]-801c:3: Abort command issued nexus=3:0:2 --  1 2002.
    qla2xxx [0000:81:00.0]-801c:3: Abort command issued nexus=3:0:2 --  1 2002.
    qla2xxx [0000:81:00.0]-8009:3: DEVICE RESET ISSUED nexus=3:0:2 cmd=ffff882f89c2c7c0.
    qla2xxx [0000:81:00.0]-800c:3: do_reset failed for cmd=ffff882f89c2c7c0.
    qla2xxx [0000:81:00.0]-800f:3: DEVICE RESET FAILED: Task management failed nexus=3:0:2 cmd=ffff882f89c2c7c0.
    qla2xxx [0000:81:00.0]-8009:3: TARGET RESET ISSUED nexus=3:0:2 cmd=ffff882f89c2c7c0.
    qla2xxx [0000:81:00.0]-800c:3: do_reset failed for cmd=ffff882f89c2c7c0.
    qla2xxx [0000:81:00.0]-800f:3: TARGET RESET FAILED: Task management failed nexus=3:0:2 cmd=ffff882f89c2c7c0.
    qla2xxx [0000:81:00.0]-8012:3: BUS RESET ISSUED nexus=3:0:2.
    qla2xxx [0000:81:00.0]-802b:3: BUS RESET SUCCEEDED nexus=3:0:2.
    qla2xxx [0000:81:00.0]-505f:3: Link is operational (8 Gbps).
    qla2xxx [0000:81:00.0]-8018:3: ADAPTER RESET ISSUED nexus=3:0:2.
    qla2xxx [0000:81:00.0]-00af:3: Performing ISP error recovery - ha=ffff88bf04d18000.
     rport-3:0-0: blocked FC remote port time out: removing target and saving binding
    qla2xxx [0000:81:00.0]-505f:3: Link is operational (8 Gbps).
    qla2xxx [0000:81:00.0]-8017:3: ADAPTER RESET SUCCEEDED nexus=3:0:2.
     rport-2:0-0: blocked FC remote port time out: removing target and saving binding
    sg_rq_end_io: device detached
    BUG: unable to handle kernel NULL pointer dereference at 00000000000002a8
    IP: [<ffffffff8133b268>] __pm_runtime_idle+0x28/0x90
    PGD 7e6586f067 PUD 7e5af06067 PMD 0 [1739975.390354] Oops: 0002 [#1] SMP
    CPU 0
    ...
    Supported: No, Proprietary modules are loaded [1739975.390463]
    Pid: 27965, comm: ABCD Tainted: PF           X 3.0.101-0.29-default #1 HP ProLiant DL580 Gen8
    RIP: 0010:[<ffffffff8133b268>]  [<ffffffff8133b268>] __pm_runtime_idle+0x28/0x90
    RSP: 0018:ffff8839dc1e7c68  EFLAGS: 00010202
    RAX: 0000000000000000 RBX: ffff883f0592fc00 RCX: 0000000000000090
    RDX: 0000000000000000 RSI: 0000000000000004 RDI: 0000000000000138
    RBP: 0000000000000138 R08: 0000000000000010 R09: ffffffff81bd39d0
    R10: 00000000000009c0 R11: ffffffff81025790 R12: 0000000000000001
    R13: ffff883022212b80 R14: 0000000000000004 R15: ffff883022212b80
    FS:  00007f8e54560720(0000) GS:ffff88407f800000(0000) knlGS:0000000000000000
    CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
    CR2: 00000000000002a8 CR3: 0000007e6ced6000 CR4: 00000000001407f0
    DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
    DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
    Process ABCD (pid: 27965, threadinfo ffff8839dc1e6000, task ffff883592e0c640)
    Stack:
     ffff883f0592fc00 00000000fffffffa 0000000000000001 ffff883022212b80
     ffff883eff772400 ffffffffa03fa309 0000000000000000 0000000000000000
     ffffffffa04003a0 ffff883f063196c0 ffff887f0379a930 ffffffff8115ea1e
    Call Trace:
     [<ffffffffa03fa309>] st_open+0x129/0x240 [st]
     [<ffffffff8115ea1e>] chrdev_open+0x13e/0x200
     [<ffffffff811588a8>] __dentry_open+0x198/0x310
     [<ffffffff81167d74>] do_last+0x1f4/0x800
     [<ffffffff81168fe9>] path_openat+0xd9/0x420
     [<ffffffff8116946c>] do_filp_open+0x4c/0xc0
     [<ffffffff8115a00f>] do_sys_open+0x17f/0x250
     [<ffffffff81468d92>] system_call_fastpath+0x16/0x1b
     [<00007f8e4f617fd0>] 0x7f8e4f617fcf
    Code: eb d3 90 48 83 ec 28 40 f6 c6 04 48 89 6c 24 08 4c 89 74 24 20 48 89 fd 48 89 1c 24 4c 89 64 24 10 41 89 f6 4c 89 6c 24 18 74 11 <f0> ff 8f 70 01 00 00 0f 94 c0 45 31 ed 84 c0 74 2b 4c 8d a5 a0
    RIP  [<ffffffff8133b268>] __pm_runtime_idle+0x28/0x90
     RSP <ffff8839dc1e7c68>
    CR2: 00000000000002a8
    
    Analysis reveals the cause of the crash to be due to STp->device
    being NULL. The pointer was NULLed via scsi_tape_put(STp) when it
    calls scsi_tape_release(). In st_open() we jump to err_out after
    scsi_block_when_processing_errors() completes and returns the
    device as offline (sdev_state was SDEV_DEL):
    
    1180 /* Open the device. Needs to take the BKL only because of incrementing the SCSI host
    1181    module count. */
    1182 static int st_open(struct inode *inode, struct file *filp)
    1183 {
    1184         int i, retval = (-EIO);
    1185         int resumed = 0;
    1186         struct scsi_tape *STp;
    1187         struct st_partstat *STps;
    1188         int dev = TAPE_NR(inode);
    1189         char *name;
    ...
    1217         if (scsi_autopm_get_device(STp->device) < 0) {
    1218                 retval = -EIO;
    1219                 goto err_out;
    1220         }
    1221         resumed = 1;
    1222         if (!scsi_block_when_processing_errors(STp->device)) {
    1223                 retval = (-ENXIO);
    1224                 goto err_out;
    1225         }
    ...
    1264  err_out:
    1265         normalize_buffer(STp->buffer);
    1266         spin_lock(&st_use_lock);
    1267         STp->in_use = 0;
    1268         spin_unlock(&st_use_lock);
    1269         scsi_tape_put(STp); <-- STp->device = 0 after this
    1270         if (resumed)
    1271                 scsi_autopm_put_device(STp->device);
    1272         return retval;
    
    The ref count for the struct scsi_tape had already been reduced
    to 1 when the .remove method of the st module had been called.
    The kref_put() in scsi_tape_put() caused scsi_tape_release()
    to be called:
    
    0266 static void scsi_tape_put(struct scsi_tape *STp)
    0267 {
    0268         struct scsi_device *sdev = STp->device;
    0269
    0270         mutex_lock(&st_ref_mutex);
    0271         kref_put(&STp->kref, scsi_tape_release); <-- calls this
    0272         scsi_device_put(sdev);
    0273         mutex_unlock(&st_ref_mutex);
    0274 }
    
    In scsi_tape_release() the struct scsi_device in the struct
    scsi_tape gets set to NULL:
    
    4273 static void scsi_tape_release(struct kref *kref)
    4274 {
    4275         struct scsi_tape *tpnt = to_scsi_tape(kref);
    4276         struct gendisk *disk = tpnt->disk;
    4277
    4278         tpnt->device = NULL; <<<---- where the dev is nulled
    4279
    4280         if (tpnt->buffer) {
    4281                 normalize_buffer(tpnt->buffer);
    4282                 kfree(tpnt->buffer->reserved_pages);
    4283                 kfree(tpnt->buffer);
    4284         }
    4285
    4286         disk->private_data = NULL;
    4287         put_disk(disk);
    4288         kfree(tpnt);
    4289         return;
    4290 }
    
    Although the problem was reported on SLES11.3 the problem appears
    in linux-next as well.
    
    The crash is fixed by reordering the code so we no longer access
    the struct scsi_tape after the kref_put() is done on it in st_open().
    Signed-off-by: NShane Seymour <shane.seymour@hp.com>
    Signed-off-by: NDarren Lavender <darren.lavender@hp.com>
    Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.com>
    Acked-by: NKai Mäkisara <kai.makisara@kolumbus.fi>
    Cc: stable@vger.kernel.org
    Signed-off-by: NJames Bottomley <JBottomley@Odin.com>
    e7ac6c66
st.c 127.9 KB