lkml.org 
[lkml]   [2004]   [Jan]   [31]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFC][PATCH] linux-2.6.2-rc2_vsyscall-gtod_B1.patch
On Wed, Jan 28, 2004 at 09:06:34PM -0800, Ulrich Drepper wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> john stultz wrote:
>
> > Please let me know if you have any comments or suggestions.
>
> I really don't like this special address in the vdso approach. Yes,
> it's unfortunately done for x86-64 as well but this doesn't mean the
> mistakes have to be repeated.

we investigated all possible implementations and we choosed for x86-64
the most efficient possible one. I think the current api was suggested
originally by hpa. Any other implementation would be running slower
period, so I wouldn't call it a mistake, I definitely call it a great
success, infact it is the best performing one for x86 too (this is why I
think john is using it). I know it's not nice from a computer science
theorical standpoint compared to other much slower implementations, but
when I run gettimeofday I want it to run as fast as possible, and I
don't care about anything else (well, besides the result being correct
of course ;), and I think the industry at large has my same needs. So I
definitely wouldn't trade it with anything else.

I'm unsure if we took care of implementing the backwards compatibility
-ENOSYS in the kernel at the next offsets of the vsyscalls, for making
it trivially extensible, if they're still missing we may want to add
them (there's no need to waste physical ram to do that btw). I had them
in my todo list for a while but at least from my part I never
implemented it, I'm sure I mentioned this had to be implemented a few
times though. Not sure if Andi or somebody else added the compatibility
-ENOSYS in the meantime. This is the sort of thing that nobody will
care about until it's too late. Well, it's not too bad anyways, the
current upgrade path would simply force an upgrade of kernel after
adding a glibc that has knowledge of the new vsyscalls, and overall
there would be no risk of any silent malfunction, it could only segfault
apps "safely". Also there is already space for at least two more
vsyscalls that currently are returning -ENOSYS. So overall even if we
don't add it, it probably won't matter much.
-
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/

\
 
 \ /
  Last update: 2005-03-22 14:00    [W:0.079 / U:2.092 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site