Message ID | cover.1742348826.git.naohiro.aota@wdc.com (mailing list archive) |
---|---|
Headers | show |
Series | btrfs: zoned: skip reporting zone for new block group | expand |
On Wed, Mar 19, 2025 at 10:49:15AM +0900, Naohiro Aota wrote: > Newly created block group should reside on empty zones, whose write pointer > should always be 0. Also, we can load the zone capacity from the block > layer. So, we don't need to REPORT_ZONE to load the info. > > The reporting zone on a new block group is not only unnecessary, but also > can cause a deadlock. When one process do a report zones and another > process freezes the block device, the report zones side cannot allocate > a tag because the freeze is already started. This can thus result in new > block group creation to hang forever, blocking the write path. > > v1: https://patch.msgid.link/cover.1741596325.git.naohiro.aota@wdc.com > v2: https://patch.msgid.link/cover.1742259006.git.naohiro.aota@wdc.com > - Move other zone related functions into the same #ifdef block. > v3: > - Rename argument variable and fix the kerneldoc. > > Naohiro Aota (2): > block: introduce zone capacity helper > btrfs: zoned: skip reporting zone for new block group There seem to be no further comments. I'll add the patches to for-next and queue it for merge in 6.15-rc2. Thanks.