https://bugs.kde.org/show_bug.cgi?id=513176
--- Comment #3 from Méven <[email protected]> --- Git commit bf22fdb72b361273987f37a9dc2001cbf9f07f85 by Méven Car. Committed on 18/10/2026 at 08:32. Pushed by meven into branch 'master'. kmountpoint: read a mount by its unique id with statmount() On a cache miss, and after any change to the mount table, currentMountPointForUniqueId() re-read the whole table with currentMountPoints(), which statx()es every mount point and lists the gvfs FUSE directory. A statx() walks the path of the mount point. When a mount point is below a network mount whose server does not answer, AT_STATX_DONT_SYNC does not stop that walk from asking the server. The lookup then blocked, even for the id of a local disk. With a tmpfs mounted inside an NFS hard mount whose server was cut off, isSlow() of a local item blocked for more than 15 s. It now takes about 1 ms. statmount() reads one mount from the kernel mount tree without walking a path. It came with Linux 6.8, like STATX_MNT_ID_UNIQUE. A kernel that does not report supported_mask keeps the table re-read. Related: bug 272361 M +26 -0 autotests/kmountpointtest.cpp M +1 -0 autotests/kmountpointtest.h M +15 -0 src/core/ConfigureChecks.cmake M +1 -0 src/core/config-kmountpoint.h.cmake M +107 -5 src/core/kmountpoint.cpp https://invent.kde.org/frameworks/kio/-/commit/bf22fdb72b361273987f37a9dc2001cbf9f07f85 -- You are receiving this mail because: You are watching all bug changes.
