Remove use on kernel spinlock and quota headers (kernel
headers are a moving target and generally unsuitable for
direct inclusion in userspace tools). Fortunately its a
straightforward change to define the couple of structures
we need here locally.
Several other code tweaks in the project quota code (for
SGI folks to review closely please):
fix the 1st arg in call to Q_XQUOTASYNC quotactl
delay doing the inode sync until we know we must, giving
plenty of opportunity to bail on XFS filesystems not using
project quota.
check return values from malloc/calloc
clear the acct/enfd flags early on, to ensure they are
not accidentally left.
short-circuit the code in the case where project quota
is not enabled, no point scanning projects file then.
remove the persistent cache loading on prjquota indom
(not clear how this helps when filesystem indom is random
in its numbering & always has been...)
simplify the pmdacache interface use as a result.
One remaining question is why was /etc/projid not used in
place of /etc/projects, to get the symbolic project names
instead of numerics?
Remove use on kernel spinlock and quota headers (kernel
headers are a moving target and generally unsuitable for
direct inclusion in userspace tools). Fortunately its a
straightforward change to define the couple of structures
we need here locally.
Several other code tweaks in the project quota code (for
SGI folks to review closely please):
One remaining question is why was /etc/projid not used in
place of /etc/projects, to get the symbolic project names
instead of numerics?