Messages in this thread | | | Date | Tue, 28 Mar 2000 07:53:25 -0600 (CST) | From | Jesse Pollard <> | Subject | Re: Avoiding OOM on overcommit...? |
| |
--------- Received message begins Here --------- Marco Colombo <marco@esi.it> > On Mon, 27 Mar 2000, Linda Walsh wrote: > > > Khimenko Victor wrote: > > > There is small difference: there are disk quotas, but there are NO memory > > > quotas :-( Per-user ones, not per-process ones I mean... > > --- > > Maybe there need to be memory quotas. Dunno. But referring to the > > file system. It is the file doing the write that returns with an error > > code. Not another random process that happened to be using disk space. Also, > > reserving 5% (for root) isn't considered to be a horrible kludge to allow > > for recovery from such situations. > > But 5% reservation is not there for recovery from FS full situation. > Ext2 guys should drop a word here, but AFAIK it's mainly for performance > reasons (it was for FFS, where it was introduced the first time: at 95% > many allocation optimizations simpy failed).
It was also made available to root to use so that syslog could continue to write to the root file system. Also, I believe pipe buffers were optionally stored in the root filesystem under tight memory conditions. Filling root with user data usually caused the system to hang trying to handle pipes and domain sockets (which syslog used...).
The reserve then allowed the root user time to log in and clean up some things before the system died. Most user processes were put to sleep.
At least this was the case under UNIX System V release 2.
------------------------------------------------------------------------- Jesse I Pollard, II Email: pollard@navo.hpc.mil
Any opinions expressed are solely my own.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |