On Tue, Nov 7, 2017 at 2:26 PM, David Disseldorp <ddiss@xxxxxxx> wrote:
> Hi Ilya,
>
> On Mon, 6 Nov 2017 16:49:30 +0100, Ilya Dryomov wrote:
>
>> - /* Only reads are allowed to a read-only device */
>> -
>> - if (op_type != OBJ_OP_READ) {
>> - if (rbd_dev->mapping.read_only) {
>> - result = -EROFS;
>> - goto err_rq;
>> - }
>> - rbd_assert(rbd_dev->spec->snap_id == CEPH_NOSNAP);
>> - }
>> + rbd_assert(op_type == OBJ_OP_READ ||
>> + rbd_dev->spec->snap_id == CEPH_NOSNAP);
>
> AFAICT, the block layer can still queue non-read requests, despite the
> block device read-only policy flag, so I think an -EROFS handler needs
> to stay.
> The following was run on mainline (e4880bc5dfb1f02b152e62a894b5c6f3e995)
> with these two patches applied:
>
> RBD mapped at: /dev/rbd/rbd/img@snap
> rapido1:/# blockdev --getro /dev/rbd/rbd/img@snap
> 1
> rapido1:/# mkfs.xfs /dev/rbd/rbd/img@snap
> [ 12.346216]
> [ 12.346216] Assertion failure in rbd_queue_workfn() at line 4028:
> [ 12.346216]
> [ 12.346216] rbd_assert(op_type == OBJ_OP_READ || rbd_dev->spec->snap_id == CEPH_NOSNAP);
> [ 12.346216]
> [ 12.347529] ------------[ cut here ]------------
> [ 12.347910] kernel BUG at drivers/block/rbd.c:4028!
> [ 12.348479] invalid opcode: 0000 [#1] SMP
> [ 12.349044] Modules linked in:
> [ 12.349499] CPU: 1 PID: 17 Comm: kworker/1:0 Not tainted 4.14.0-rc8+ #405
> [ 12.350407] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-prebuilt.qemu-project.org 04/01/2014
> [ 12.352001] Workqueue: rbd rbd_queue_workfn
> [ 12.352498] task: ffff88001d27b480 task.stack: ffffc90000090000
> [ 12.353058] RIP: 0010:rbd_queue_workfn+0x4f9/0x500
> [ 12.353527] RSP: 0018:ffffc90000093e08 EFLAGS: 00010282
> [ 12.354036] RAX: 0000000000000086 RBX: ffff88001a7c7980 RCX: ffffffff81829398
> [ 12.354890] RDX: 0000000000000001 RSI: 0000000000000086 RDI: ffffffff819f9f6c
> [ 12.355580] RBP: ffffc90000093e60 R08: 00000000fffffffe R09: 000000000000048f
> [ 12.356285] R10: 0000000000000005 R11: 000000000000048e R12: ffff88001ca6f000
> [ 12.356964] R13: 0000000000000000 R14: 0000000000000002 R15: ffff88001a7c7ac0
> [ 12.357650] FS: 0000000000000000(0000) GS:ffff88001e500000(0000) knlGS:0000000000000000
> [ 12.358443] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 12.359031] CR2: 00007fcd19a14000 CR3: 000000001f069000 CR4: 00000000000006a0
> [ 12.359806] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> [ 12.360604] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> [ 12.361397] Call Trace:
> [ 12.361686] process_one_work+0x11e/0x280
> [ 12.362145] worker_thread+0x45/0x3b0
> [ 12.362564] kthread+0xff/0x140
> [ 12.362935] ? process_one_work+0x280/0x280
> [ 12.363409] ? kthread_create_on_node+0x40/0x40
> [ 12.363936] ret_from_fork+0x22/0x30
> [ 12.364338] Code: 85 e1 fb ff ff e9 99 fb ff ff 48 c7 c1 e8 af 71 81 ba bc 0f 00 00 48 c7 c6 f0 bb 64 81 48 c7 c7 40 9d 71 81 31 c0 e8 65 df d5 ff <0f> 0b 0f 1f 44 00 00 55 48 89 e5 41 56 41 55 41 54 49 89 fc 53
> [ 12.366440] RIP: rbd_queue_workfn+0x4f9/0x500 RSP: ffffc90000093e08
> [ 12.367147] ---[ end trace 75c22000d1a0b248 ]---
Hi David,
Heh, apparently I only tested writes and didn't do discards... Writes
go through blkdev_write_iter() but zeroout/discards seem to rely solely
on FMODE_WRITE checks, which the block layer doesn't enforce.
I'd rather try to fix it in the block layer before working around it in
krbd. I'll work up a patch tomorrow.
Thanks,
Ilya
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
