Hi all, I run into a problem, that is when I mapped a nbd device with rbd-nbd map CLI, the reproduce steps are: 1. rbd-nbd map test # 'test' size is 5G 2. mkfs.ext4 /dev/nbd0 # nbd0 is the mapped device 3. mount /dev/nbd0 /mnt # 4. rbd resize test 6G # resize 'test' to 6G in ceph cluster 5. lsblk # nbd0 has new size 6G 6. ls /mnt # the ls proc will hang forever, 'kill -9' on it cann't kill it. the stack of ls proc is: [<ffffffff8d1304ce>] radix_tree_lookup_slot+0x1e/0x50 [<ffffffff8cf7d27b>] find_get_entry+0x1b/0x100 [<ffffffff8cf7d750>] pagecache_get_page+0x30/0x2b0 [<ffffffff8d1304ce>] radix_tree_lookup_slot+0x1e/0x50 [<ffffffff8cf7d27b>] find_get_entry+0x1b/0x100 [<ffffffff8cf7d750>] pagecache_get_page+0x30/0x2b0 [<ffffffff8d03b7a1>] __getblk_gfp+0x101/0x310 [<ffffffffc0586818>] ext4_getblk+0xa8/0x190 [ext4] [<ffffffffc058691f>] ext4_bread+0x1f/0xb0 [ext4] [<ffffffff8d01be68>] dput+0x38/0x250 [<ffffffffc0590680>] __ext4_read_dirblock+0x30/0x350 [ext4] [<ffffffffc057ec15>] ext4_file_open+0xa5/0x270 [ext4] [<ffffffffc0590fb9>] htree_dirblock_to_tree+0x69/0x2b0 [ext4] [<ffffffff8cf85f48>] get_page_from_freelist+0x908/0xb40 [<ffffffff8d00d8c9>] terminate_walk+0x89/0xf0 [<ffffffffc05921f6>] ext4_htree_fill_tree+0xa6/0x2e0 [ext4] [<ffffffff8cf87206>] __alloc_pages_nodemask+0xf6/0x260 [<ffffffff8cfe147b>] kmem_cache_alloc_trace+0x11b/0x530 [<ffffffffc057e5b2>] ext4_readdir+0x6f2/0xa20 [ext4] [<ffffffff8d01819b>] iterate_dir+0x16b/0x190 [<ffffffff8d018688>] SyS_getdents+0x98/0x120 [<ffffffff8d0183f0>] fillonedir+0xe0/0xe0 [<ffffffff8d40861e>] system_call_fast_compare_end+0xc/0xb7 [<ffffffffffffffff>] 0xffffffffffffffff and I can see an error log in rbd-nbd client log file: 2018-11-28 14:30:19.020284 7f7aa27fc700 -1 rbd-nbd: rescan of partition table failed: (16) Device or resource busy If I don't mount nbd0 and run resize2fs on nbd0, all things ok, then I mount nbd0 after resize2fs is ok, too. If I map test vol with 'rbd map test' CLI, then the steps 1~6 above are all ok, the ext4 fs on rbd0 is resized correctly. If I use xfs filesystem on nbd0 with rbd-nbd mapped, the steps 1~6 above are all ok with a resized filesystem, even there is an error log in rbd-nbd client log file. The ceph version is 12.2.5, with kernel 4.9.65. Thanks a lot for any help.
