[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
Re: [patches] patch-2.2.2-nonx86,patch-2.2.2-filter
- From: Michal Jaegermann <michal ellpspace math ualberta ca>
- To: rbowles infonautics com
- Cc: axp-list redhat com
- Subject: Re: [patches] patch-2.2.2-nonx86,patch-2.2.2-filter
- Date: Tue, 23 Feb 1999 23:26:32 -0700 (MST)
Robert Bowles wrote:
>
> Changes/Observations
> - Socket Filtering Code
> This alone should be a showstopper.
> I don't see where this could work on any platform.
> 'struct sk_filter' is referenced but not defined _anywhere_.
> The only solution is to build w/o socket filtering code.
Patch for that was already postes five or six times (maybe more,
I lost count) on linux-kernel list. Half of an original patch
was lost in a shuffle. 2.2.2-ac1 has this few lines already applied.
> - Semaphore Code
> All seems well, except for the whole new
> '__down_trylock() / waking_non_zero_trylock()' bit.
This bit is for the moment "Intel only". Linus says that he
did that on purpose to force the issue and get people to bring
other platforms in sync instead of causing some subtle breakage.
Daniel J. Frasnelli posted a temporary patch for that already;
both on axp-list and on linux-kernel list. And also actually
a few times.
It is quite likely that the patch causes this "subtle breakage" mentioned
above as it backs off __down_trylock() and some races may show up -
infrequently. I am running 2.2.2-ac1 patched that way right now.
Cross your fingers. :-)
> On the minus, socket filtering is broken.
Yes, I did compile with socket filtering on - just to be sure.
But this is 'ac1' and not "pure" 2.2.2.
Michal
[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
[]