System might be under heavy I/O pressure. If it takes long time to get data from disk we should load from network instead.
I am a little bit skeptical about this. Fetching from the network in general is very slow (outside simulated tests) and bad for battery life.
Created attachment 258192 [details]
(In reply to comment #1)
> I am a little bit skeptical about this. Fetching from the network in general
> is very slow (outside simulated tests) and bad for battery life.
We have spintraces where system is crawling because heavy I/O load (generally systems with spinning disks). It is better to keep the browser functional. The timeout is very long (1s) should not get hit under any normal circumstances.
Comment on attachment 258192 [details]
View in context: https://bugs.webkit.org/attachment.cgi?id=258192&action=review
At first I was skeptical. But 1 second timeout on disk read -- okay, we're pretty damn hosed.
> + This patch introduces 1s second timeout for disk reads. If the last dispatched cache read takes longer than that we cancel all
"1s second timeout"
http://trac.webkit.org/changeset/187958 (with 1.5s timeout)