1341247 - Firefox would be blocked by mozilla::net::_OldStorage::AsyncEvictStorage()

Status: RESOLVED INVALID

(Core :: Networking: Cache, defect)

Description

[Thinker Li [:sinker]]

For some reason, Firefox starts to evict cache storage and block the main thread of the chrome process for minutes.  I studied the behavior of the chrome process of Firefox with strace on Linux, it spent a lot of time on accessing cache files.  With gdb attaching on the chrome process, it stayed in mozilla::net::_OldStorage::AsyncEvictStorage()for long at the main thread.

Once the main thread of the chrome process is blocked, user see black on the screen.

Comment 1

[Honza Bambas (:mayhemer)]

Do you have the whole stack?  What was the "remember history" setting in that profile?

Comment 2

[Thinker Li [:sinker]]

Following link is the backtrace.
  https://public.etherpad-mozilla.org/p/7OrDIpacJf

The setting of history is "Remember history".
I am using nightly build.
Do you need more information?

Comment 3

[Honza Bambas (:mayhemer)]

(In reply to Thinker Li [:sinker] from comment #2)
> Following link is the backtrace.
>   https://public.etherpad-mozilla.org/p/7OrDIpacJf
> 
> The setting of history is "Remember history".
> I am using nightly build.
> Do you need more information?

There are no line numbers, but it seems to be invoked from [1] where are accessing each cache storage via the new cache2 API.  Check your preferences are:

browser.cache.use_new_backend_temp;true
browser.cache.use_new_backend;1

because you are apparently not using the cache2 back end, but the old one that can only be enabled via these prefs.

[1] https://dxr.mozilla.org/mozilla-central/rev/a180b976c165b6cd174d24bbb77941919cdc53cb/netwerk/cache2/CacheObserver.cpp#562

Comment 4

[Thinker Li [:sinker]]

Hi Honza,

Thank you!  My preference is just opposite to your suggestion.  They are false and 0.  I will try it for awhile to see what would happen.