Re: [PATCH] Limit file descriptors used by packs
- From
Shawn O. Pearce <spearce@spearce.org>
- Date
- Feb 28, 2011, 20:44 UTC
- Message-ID
- <20110228204406.GA26052@spearce.org>
- In-Reply-To
- <20110228203557.GA8189@pcpool00.mathematik.uni-freiburg.de>
"Bernhard R. Link" <brl+ccmadness@pcpool00.mathematik.uni-freiburg.de> wrote:
Show 10 quoted lines
> * Shawn O. Pearce <spearce@spearce.org> [110228 21:27]: > > By using a hard upper limit that is below the rlimit of the current > > process, it is not necessary to check for EMFILE on every single > > fd-allocating system call. Instead reserving 8 file descriptors > > makes it safe to assume the system call won't fail due to being > > over limit in the filedescriptor limit. > > Isn't 8 quite a bit low for a reserve? Couldn't some libc stuff > (especially nss modules perhaps activated by something) easily surpass > that?
Originally I proposed 25 to Junio, but he scoffed and said that was quite high. So I went with 8, 3 for std{in,out,err} and 5 as a WAG for everything else.
Its arbitrary, 25 might be a better WAG than 8...
-- Shawn.