Hi David, Jeffle & Xiang In production environment, process crashes sometimes occurs. In cachefiles on-demand read scenario, if user daemon crashes, requests will return -EIO. User programs which do not consider this error will trap into uncertain state. Based on this, we came up with a user daemon crash recover scheme. Even if user daemon crashes, the device connection and anonymous fd will not be released. Recovered user daemon only needs to write 'restore' to /dev/cachefiles to restore in-flight requests. Userspace Crash Recover Demo (Based on Jeffle's User Demo) -------------------------- Git tree: https://github.com/userzj/demand-read-cachefilesd.git main Gitweb: https://github.com/userzj/demand-read-cachefilesd Jia Zhu (1): cachefiles: Add a command to restore on-demand requests fs/cachefiles/daemon.c | 1 + fs/cachefiles/internal.h | 3 +++ fs/cachefiles/ondemand.c | 25 +++++++++++++++++++++++++ 3 files changed, 29 insertions(+) -- 2.20.1 -- Linux-cachefs mailing list Linux-cachefs@xxxxxxxxxx https://listman.redhat.com/mailman/listinfo/linux-cachefs
