Re: [PATCH] log: setlogmask(0) can actually result in a system call, don't use it

[Stefano Brivio <sbrivio@redhat.com>]

On Fri, 2 Feb 2024 12:02:19 +1100
David Gibson <david@gibson.dropbear.id.au> wrote:

> On Fri, Feb 02, 2024 at 12:32:57AM +0100, Stefano Brivio wrote:
> > Before commit 32d07f5e59f2 ("passt, pasta: Completely avoid dynamic
> > memory allocation"), we didn't store the current log mask in a
> > variable, and we fetched it using setlogmask(0) wherever needed.
> > 
> > But after that commit, we can just use our log_mask copy instead. And
> > we should: with recent glibc versions, setlogmask(0) actually results
> > in a system call, which causes a substantial overhead with high
> > transfer rates: we use setlogmask(0) even to decide we don't want to
> > print debug messages.
> > 
> > Signed-off-by: Stefano Brivio <sbrivio@redhat.com>  
> 
> The basic idea looks fine to me, but there's a curly edge case.  At
> present vlogmsg() works ok before __openlog(), sending the message to
> stderr.  We rely on this because there are die() calls before our
> __openlog(), which go via err() -> vlogmsg().
> 
> With this change, before the first __setlogmask(), which occurs right
> after the __openlog() then log_mask is initialized to zero.  Which
> means, AFAICT, that we'll suppress all error messages before that
> point, which isn't great.

Oops, right, thanks for pointing this out. I had a vague memory of some
side benefit of setlogmask(0) with this regard but I couldn't really
remember and gave up on it -- it was exactly this.

I'm re-posting this with a static initialisation of log_mask now.

-- 
Stefano