I've started looking at how we could fix the following bug:
https://bugzilla.redhat.com/show_bug.cgi?id=1786222
Long story short, we are not using just async signal safe functions in a
forked off thread that runs chown() from within the domain's private
namespace. Hence, we ran into a deadlock (although the case of the bug
the deadlock is in glibc not gluster).
So far I don't have any fix, but I have couple of cleanups.
Michal Prívozník (8):
security_dac: Use g_autofree
security_dac: Introduce virSecurityDACChownItemFree()
security_dac: Introduce g_autoptr for virSecurityDACChownList
security_dac: Don't check for !priv in
virSecurityDACSetOwnershipInternal()
virSecurityDACSetOwnershipInternal: Drop dead code
virSecurityDACSetOwnershipInternal: Don't overwrite @path argument
virSecurityDACSetOwnershipInternal: Fix WIN32 code
qemu: Deduplicate code in qemuSecurityChownCallback()
src/qemu/qemu_driver.c | 22 +---
src/security/security_dac.c | 201 ++++++++++++++------------------
src/security/security_manager.h | 13 ++-
3 files changed, 98 insertions(+), 138 deletions(-)
--
2.31.1