https://bugzilla.kernel.org/show_bug.cgi?id=198187 --- Comment #12 from Mike Snitzer (snitzer@xxxxxxxxxx) --- (In reply to Jan Kara from comment #11) > Thanks. Mike, any idea here? It appears JBD2 submitted IO to dm-raid1 and it > never completed. Apparently this did not happen with 4.12.8 but it does > happen relatively regularly with 4.13.8... git log --oneline v4.12.8^..v4.13.8 -- drivers/md/dm-raid1.c | tac 9966afa dm: fix REQ_RAHEAD handling 846785e dm: don't return errnos from ->map 1be5690 dm: change ->end_io calling convention 4e4cbee block: switch bios to blk_status_t So just changes from Christoph to prepare for and finally switch to blk_status_t $ git log --oneline v4.12.8^..v4.13.8 -- drivers/md/dm.c | tac 846785e dm: don't return errnos from ->map 1be5690 dm: change ->end_io calling convention 4e4cbee block: switch bios to blk_status_t 7e026c8 dm: add ->copy_from_iter() dax operation support abebfbe dm: add ->flush() dax operation support 011067b blk: replace bioset_create_nobvec() with a flags arg to bioset_create() 47e0fb46 blk: make the bioset rescue_workqueue optional. 93e6442 dm: add basic support for using the select or poll function a4aa5e5 dm: fix REQ_OP_ZONE_RESET bio handling 264c869 dm: fix REQ_OP_ZONE_REPORT bio handling 1099930 dm: introduce dm_remap_zone_report() 41341af dm: don't set bounce limit fbd08e7 bio-integrity: fix interface for bio_integrity_trim 54385bf dm: fix the second dec_pending() argument in __split_and_process_bio() 6044078 dm: fix printk() rate limiting code bfc0ab4 dax: remove the pmem_dax_ops->flush abstraction 666cb84 dm ioctl: fix alignment of event number in the device list Nothing in these DM core changes stand out to me as a potential source for IO hangs. But obviously the devil is in the details... Could easily be something in block core is negatively influencing DM, etc. -- You are receiving this mail because: You are watching the assignee of the bug.
