| Age | Commit message (Collapse) | Author |
|---|
|
a hard limit of 2x. add a comment to this effect. reviewed by miod
|
|
restrict lock count per uid to a global limit, add sysctl to adjust limit.
this prevents a user from creating too many locks. problem noticed
by devon o'dell. ok deraadt miod pedro
|
|
|
|
From NetBSD.
Tested by many people, ok art@.
|
|
takes a void *. convert uiomove to take a void * as well. ok deraadt@
|
|
rescinded 22 July 1999. Proofed by myself and Theo.
|
|
|
|
costa@ ok.
|
|
but not wakeup'ed. In this case one entry can be placed twice at list of
blocked locks. As a result block list contain an entry which points to
itself. lf_wakelock can't remove such an entry and system livelocks
trying to remove a bad entry from block list.
Other changes include:
- repair debug functions and make vfs_lockf.c compilable with LOCKF_DEBUG
- simplify debug printing and remove useless debugging
- remove unnecessary casting, replace NOLOCKF with NULL
- free -> FREE (use one form over the file)
- convert list of blocked locks to use TAILQ_* macroses
- lf_addblock() -> TAILQ_INSERT_TAIL
- Fix bug in lf_findoverlap(): in old code
if (end == -1) && (lf->lf_end == -1) && (lf->lf_start <= start)
then lf_findoverlap() return 4 instead of 2
- work more carefully with pointers (probably fix one or two bugs)
- use wakeup_one()
- KNF
|
|
|
|
|
