Keep clone files and oversized responses out of the cache

The dumb transport reads files that already sit on the disk, so a pack
copied into a slot cost that disk twice and the request a second write
of every byte. A snapshot took a slot whatever its size, so a visitor
naming distinct refs and ids could fill the cache root with archives.
`cache-max-slot-size`, 64 MB unless set, now serves a larger response
from the lock file and drops it, along with any expired copy it would
have replaced.

1 file changed, 6 insertions, 2 deletions

diff --git a/source/cache.h b/source/cache.h
index bae283c..d3a7ec4 100644
--- a/source/cache.h
+++ b/source/cache.h
@@ -7,6 +7,8 @@
 #ifndef CGIT_CACHE_H
 #define CGIT_CACHE_H
 
+#include <stddef.h>
+
 typedef void (*cache_fill_fn)(void);
 
 /*
@@ -14,10 +16,12 @@ typedef void (*cache_fill_fn)(void);
  * slot for that key is there and rendering it through fn when it is not. size
  * is how many slots the cache may use and path is the directory holding them.
  * ttl is how many minutes a slot for this key stays fresh, where a negative
- * ttl never expires and a ttl of zero skips the cache for this request.
+ * ttl never expires and a ttl of zero skips the cache for this request. A
+ * page larger than max_bytes is served but not kept, and zero sets no bound.
  * Returns 0 when the page was written, and an errno value when it was not.
  */
-extern int cache_process(int size, const char *path, const char *key, int ttl, cache_fill_fn fn);
+extern int cache_process(int size, const char *path, const char *key, int ttl,
+	size_t max_bytes, cache_fill_fn fn);
 
 /*
  * Write one line per cache slot to stdout, giving its path, modification time,