3d9db48dbe
Previously, we would read the entire padding in a given archive into memory in order to store it in the packer. This would cause memory exhaustion if a malicious archive was crafted with very large amounts of padding. Since a given SegmentType is reconstructed losslessly, we can simply chunk up any padding into large segments to avoid this problem. Use a reasonable default of 1MiB to avoid changing the tar-split.json of existing archives that are not malformed. Fixes: CVE-2017-14992 Signed-off-by: Aleksa Sarai <asarai@suse.de> |
||
---|---|---|
.. | ||
testdata | ||
assemble.go | ||
assemble_test.go | ||
disassemble.go | ||
doc.go | ||
README.md |
asm
This library for assembly and disassembly of tar archives, facilitated by
github.com/vbatts/tar-split/tar/storage
.
Concerns
For completely safe assembly/disassembly, there will need to be a Content
Addressable Storage (CAS) directory, that maps to a checksum in the
storage.Entity
of storage.FileType
.
This is due to the fact that tar archives can allow multiple records for the same path, but the last one effectively wins. Even if the prior records had a different payload.
In this way, when assembling an archive from relative paths, if the archive has multiple entries for the same path, then all payloads read in from a relative path would be identical.
Thoughts
Have a look-aside directory or storage. This way when a clobbering record is encountered from the tar stream, then the payload of the prior/existing file is stored to the CAS. This way the clobbering record's file payload can be extracted, but we'll have preserved the payload needed to reassemble a precise tar archive.
clobbered/path/to/file.[0-N]
alternatively
We could just not support tar streams that have clobbering file paths. Appending records to the archive is not incredibly common, and doesn't happen by default for most implementations. Not supporting them wouldn't be a security concern either, as if it did occur, we would reassemble an archive that doesn't validate signature/checksum, so it shouldn't be trusted anyway.
Otherwise, this will allow us to defer support for appended files as a FUTURE FEATURE.