On Sunday 24 March 2002 20:50, Alan Cox wrote:
> > How does a process react that doesn't get no more memory?
>
> Thats up to the process. If a program doesn't handle malloc/mmap/etc
> failures then its junk anyway
Agreed 101%.
On Sunday 24 March 2002 21:43, Paul P Komkoff Jr wrote:
> Replying to Alan Cox:
> > Thats up to the process. If a program doesn't handle malloc/mmap/etc
> > failures then its junk anyway
>
> The recent junk I fighting with to take full advantage of overcommit
> accounting is squid.
> Very popular junk. Maybe rsync uses same 'secret technique' to handle
> malloc failures? :)))
>
> Btw. Overcommit handling not very good yet.
> Squid hits the limit, then bails out. Then shell script trying to start new
> instance of squid (actually trying to sleep before restart), but gets
> 'fork - cannot allocate memory'. seems that memory isn't dealloced
> from already exited process space :(
In all the code I've written I always check if malloc() and friends don't
succeed. In some instances the code can even handle this gracefully and
malloc failure doesn't cause a fatal application error.
Now if kernel kills my process it doesn't matter if I can handle the out of
memory condition or not. The kernel already made a decision for me instead of
my application... and never mind the data I have cached in applications
buffers... the kernel also decided it wasn't worth writing to a disk.
--------------------------------------------------------------------------
Alan.. I would find it acceptable if a process is killed if it tries to
malloc() memory *after* a previous malloc has already failed and there is
still out-of-memory condition.
-- best regards, Rok Papeľ. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/