]> git-server-git.apps.pok.os.sepia.ceph.com Git - ceph.git/commit
librbd/cache/pwl/ssd: solve competition between read and retire 42580/head
authorFeng Hualong <hualong.feng@intel.com>
Fri, 30 Jul 2021 12:24:05 +0000 (20:24 +0800)
committerIlya Dryomov <idryomov@gmail.com>
Wed, 11 Aug 2021 20:12:10 +0000 (22:12 +0200)
commitdfdb452aa996af40f7b0ec684670ccf9a9b2d4c1
tree21538d5424c41d9fa75c62dbf1e9e7163b04a426
parent0dedabe15a9aababb86f9a9a1c0b9b0e94e0729b
librbd/cache/pwl/ssd: solve competition between read and retire

SSD read is not like rwl's. SSD need aio read. Therefore,
we cannot guarantee that the data will not be retire
during the period from sending the read request to the SSD
and receiving the data to the memory, which may cause
the corresponding data on the SSD to be overwritten.

Fixes: https://tracker.ceph.com/issues/52236
Signed-off-by: Feng Hualong <hualong.feng@intel.com>
src/librbd/cache/pwl/AbstractWriteLog.cc
src/librbd/cache/pwl/AbstractWriteLog.h
src/librbd/cache/pwl/LogEntry.cc
src/librbd/cache/pwl/LogEntry.h
src/librbd/cache/pwl/rwl/LogEntry.cc
src/librbd/cache/pwl/rwl/LogEntry.h
src/librbd/cache/pwl/rwl/WriteLog.cc
src/librbd/cache/pwl/rwl/WriteLog.h
src/librbd/cache/pwl/ssd/LogEntry.h
src/librbd/cache/pwl/ssd/WriteLog.cc
src/librbd/cache/pwl/ssd/WriteLog.h