fbaf1171b4
An embarrassing thinko in cgit_check_cache() would truncate valid cachefiles
in the following situation:
1) process A notices a missing/expired cachefile
2) process B gets scheduled, locks, fills and unlocks the cachefile
3) process A gets scheduled, locks the cachefile, notices that the cachefile
now exist/is not expired anymore, and continues to overwrite it with an
empty lockfile.
Thanks to Linus for noticing (again).
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Cache algorithm
===============
Cgit normally returns cached pages when invoked. If there is no cache file, or
the cache file has expired, it is regenerated. Finally, the cache file is
printed on stdout.
When it is decided that a cache file needs to be regenerated, an attempt is
made to create a corresponding lockfile. If this fails, the process gives up
and uses the expired cache file instead.
When there is no cache file for a request, an attempt is made to create a
corresponding lockfile. If this fails, the process calls sched_yield(2) before
restarting the request handling.
In pseudocode:
name = generate_cache_name(request);
top:
if (!exists(name)) {
if (lock_cache(name)) {
generate_cache(request, name);
unlock_cache(name);
} else {
sched_yield();
goto top;
}
} else if (expired(name)) {
if (lock_cache(name)) {
generate_cache(request, name);
unlock_cache(name);
}
}
print_file(name);
The following options can be set in /etc/cgitrc to control cache behaviour:
cache-root: root directory for cache files
cache-root-ttl: TTL for the repo listing page
cache-repo-ttl: TTL for any repos summary page
cache-dynamic-ttl: TTL for pages with symbolic references (not SHA1)
cache-static-ttl: TTL for pages with sha1 references
TTL is specified in minutes, -1 meaning "infinite caching".
Naming of cache files
---------------------
Repository listing: <cachedir>/index.html
Repository summary: <cachedir>/<repo>/index.html
Repository subpage: <cachedir>/<repo>/<page>/<querystring>.html
The corresponding lock files have a ".lock" suffix.