[PATCH 5/5] nvme: enable logical block size > PAGE_SIZE

[Hannes Reinecke]

On 5/13/24 23:05, Luis Chamberlain wrote:
> On Mon, May 13, 2024 at 06:07:55PM +0200, Hannes Reinecke wrote:
>> On 5/12/24 11:16, Luis Chamberlain wrote:
>>> On Sat, May 11, 2024 at 07:43:26PM -0700, Luis Chamberlain wrote:
>>>> I'll try next going above 512 KiB.
>>>
>>> At 1 MiB NVMe LBA format we crash with the BUG_ON(sectors <= 0) on bio_split().
>>>
>>> [   13.401651] ------------[ cut here ]------------
>>> [   13.403298] kernel BUG at block/bio.c:1626!
>> Ah. MAX_BUFS_PER_PAGE getting in the way.
>>
>> Can you test with the attached patch?
> 
> Nope same crash:
> 
> I've enabled you to easily test with with NVMe on libvirt with kdevops,
> please test.
> 
>   Luis
> 
> [   14.972734] ------------[ cut here ]------------
> [   14.974731] kernel BUG at block/bio.c:1626!
> [   14.976906] invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
> [   14.978899] CPU: 3 PID: 59 Comm: kworker/u36:0 Not tainted 6.9.0-rc6+ #4
> [   14.981005] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
> [   14.983782] Workqueue: nvme-wq nvme_scan_work [nvme_core]
> [   14.985431] RIP: 0010:bio_split+0xd5/0xf0
> [   14.986627] Code: 5b 4c 89 e0 5d 41 5c 41 5d c3 cc cc cc cc c7 43 28 00 00 00 00 eb db 0f 0b 45 31 e4 5b 5d 4c 89 e0 41 5c 41 5d c3 cc cc cc cc <0f> 0b 0f 0b 4c 89 e7 e8 bf ee ff ff eb e1 66 66 2e 0f 1f 84 00 00
> [   14.992063] RSP: 0018:ffffbecc002378d0 EFLAGS: 00010246
> [   14.993416] RAX: 0000000000000001 RBX: ffff9e2fe8583e40 RCX: ffff9e2fdcb73060
> [   14.995181] RDX: 0000000000000c00 RSI: 0000000000000000 RDI: ffff9e2fe8583e40
> [   14.996960] RBP: 0000000000000000 R08: 0000000000000080 R09: 0000000000000000
> [   14.998715] R10: ffff9e2fe8583e40 R11: ffff9e2fe8583eb8 R12: ffff9e2fe884b750
> [   15.000510] R13: 0000000000000000 R14: 0000000000000001 R15: 0000000000000000
> [   15.002128] FS:  0000000000000000(0000) GS:ffff9e303bcc0000(0000) knlGS:0000000000000000
> [   15.003956] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [   15.005294] CR2: 0000561b2b5ce478 CR3: 0000000102484002 CR4: 0000000000770ef0
> [   15.006921] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> [   15.008509] DR3: 0000000000000000 DR6: 00000000fffe07f0 DR7: 0000000000000400
> [   15.010001] PKRU: 55555554
> [   15.010672] Call Trace:
> [   15.011297]  <TASK>
> [   15.011868]  ? die+0x32/0x80
> [   15.012572]  ? do_trap+0xd9/0x100
> [   15.013306]  ? bio_split+0xd5/0xf0
> [   15.014051]  ? do_error_trap+0x6a/0x90
> [   15.014854]  ? bio_split+0xd5/0xf0
> [   15.015597]  ? exc_invalid_op+0x4c/0x60
> [   15.016419]  ? bio_split+0xd5/0xf0
> [   15.017113]  ? asm_exc_invalid_op+0x16/0x20
> [   15.017932]  ? bio_split+0xd5/0xf0
> [   15.018624]  __bio_split_to_limits+0x90/0x2d0
> [   15.019474]  blk_mq_submit_bio+0x111/0x6a0
> [   15.020280]  ? kmem_cache_alloc+0x254/0x2e0
> [   15.021040]  submit_bio_noacct_nocheck+0x2f1/0x3d0
> [   15.021893]  ? submit_bio_noacct+0x42/0x5b0
> [   15.022658]  block_read_full_folio+0x2b7/0x350
> [   15.023457]  ? __pfx_blkdev_get_block+0x10/0x10
> [   15.024284]  ? __pfx_blkdev_read_folio+0x10/0x10
> [   15.025073]  ? __pfx_blkdev_read_folio+0x10/0x10
> [   15.025851]  filemap_read_folio+0x32/0xb0
> [   15.026540]  do_read_cache_folio+0x108/0x200
> [   15.027271]  ? __pfx_adfspart_check_ICS+0x10/0x10
> [   15.028066]  read_part_sector+0x32/0xe0
> [   15.028701]  adfspart_check_ICS+0x32/0x480
> [   15.029334]  ? snprintf+0x49/0x70
> [   15.029875]  ? __pfx_adfspart_check_ICS+0x10/0x10
> [   15.030592]  bdev_disk_changed+0x2a2/0x6e0
> [   15.031226]  blkdev_get_whole+0x5f/0xa0
> [   15.031827]  bdev_open+0x201/0x3c0
> [   15.032360]  bdev_file_open_by_dev+0xb5/0x110
> [   15.032990]  disk_scan_partitions+0x65/0xe0
> [   15.033598]  device_add_disk+0x3e0/0x3f0
> [   15.034172]  nvme_scan_ns+0x5f0/0xe50 [nvme_core]
> [   15.034862]  nvme_scan_work+0x26f/0x5a0 [nvme_core]
> [   15.035568]  process_one_work+0x189/0x3b0
> [   15.036168]  worker_thread+0x273/0x390
> [   15.036713]  ? __pfx_worker_thread+0x10/0x10
> [   15.037312]  kthread+0xda/0x110
> [   15.037779]  ? __pfx_kthread+0x10/0x10
> [   15.038316]  ret_from_fork+0x2d/0x50
> [   15.038829]  ? __pfx_kthread+0x10/0x10
> [   15.039364]  ret_from_fork_asm+0x1a/0x30
> [   15.039924]  </TASK>
> 

Ah. So this should fix it:

diff --git a/block/blk-merge.c b/block/blk-merge.c
index 4e3483a16b75..4fac11edd0c8 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -289,7 +289,7 @@ struct bio *bio_split_rw(struct bio *bio, const 
struct queue_limits *lim,

                 if (nsegs < lim->max_segments &&
                     bytes + bv.bv_len <= max_bytes &&
-                   bv.bv_offset + bv.bv_len <= PAGE_SIZE) {
+                   bv.bv_offset + bv.bv_len <= lim->max_segment_size) {
                         nsegs++;
                         bytes += bv.bv_len;
                 } else {

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                  Kernel Storage Architect
hare at suse.de                                +49 911 74053 688
SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg
HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich