diff options
| author | Tim Kientzle <kientzle@acm.org> | 2015-01-10 12:24:58 -0800 |
|---|---|---|
| committer | Tim Kientzle <kientzle@acm.org> | 2015-01-10 12:24:58 -0800 |
| commit | 6e06b1c89dd0d16f74894eac4cfc1327a06ee4a0 (patch) | |
| tree | 96e98d0fba702bdca504274db804fd153ab1448a /libarchive/archive_read.c | |
| parent | 48b288a03347e49f2f9501f040f626f916195de6 (diff) | |
| download | libarchive-6e06b1c89dd0d16f74894eac4cfc1327a06ee4a0.tar.gz | |
Fix a potential crash issue discovered by Alexander Cherepanov:
It seems bsdtar automatically handles stacked compression. This is a
nice feature but it could be problematic when it's completely
unlimited. Most clearly it's illustrated with quines:
$ curl -sRO http://www.maximumcompression.com/selfgz.gz
$ (ulimit -v 10000000 && bsdtar -tvf selfgz.gz)
bsdtar: Error opening archive: Can't allocate data for gzip decompression
Without ulimit, bsdtar will eat all available memory. This could also
be a problem for other applications using libarchive.
Diffstat (limited to 'libarchive/archive_read.c')
| -rw-r--r-- | libarchive/archive_read.c | 7 |
1 files changed, 5 insertions, 2 deletions
diff --git a/libarchive/archive_read.c b/libarchive/archive_read.c index 02bf8d3a..8f71a8b9 100644 --- a/libarchive/archive_read.c +++ b/libarchive/archive_read.c @@ -548,13 +548,13 @@ archive_read_open1(struct archive *_a) static int choose_filters(struct archive_read *a) { - int number_bidders, i, bid, best_bid; + int number_bidders, i, bid, best_bid, n; struct archive_read_filter_bidder *bidder, *best_bidder; struct archive_read_filter *filter; ssize_t avail; int r; - for (;;) { + for (n = 0; n < 25; ++n) { number_bidders = sizeof(a->bidders) / sizeof(a->bidders[0]); best_bid = 0; @@ -600,6 +600,9 @@ choose_filters(struct archive_read *a) return (ARCHIVE_FATAL); } } + archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, + "Input requires too many filters for decoding"); + return (ARCHIVE_FATAL); } /* |