From: Junio C Hamano <gitster@pobox.com>
Date: Thu, 22 Sep 2016 18:21:18 GMT
Subject: Re: [PATCH] verify_packfile: check pack validity before accessing data
Message-ID: <xmqq37kr6bj5.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <20160922040523.x3pihbs7f5iudyfz@sigill.intra.peff.net>

Jeff King <peff@peff.net> writes:

> So I wanted to know whether there were any code paths that failed to do
> so, and just blindly rely on the lazy-open. Finding the races is
> inherently hard, because you only catch them when somebody else is doing
> a repack. But if we just _remove_ the lazy-load, then it becomes easy to
> catch anybody relying on it. Like:
> ...

Clever; I like it.

> In such a case, we are relying on the lazy-load (and we _are_ racy!).
> But the patch above would punish people on low-descriptor systems. It's
> better to have an unlikely race and complete the request than to fail
> consistently. :-/
>
> For people who are running high-traffic servers, they just need to make
> sure their file descriptor limit is reasonably high to avoid the race.

Thanks for an illuminating backstory for the patch.