From: Patrick Steinhardt Date: Wed, 24 Jun 2026 09:46:27 GMT Subject: Re: [PATCH 11/11] reftable/table: fix OOB read on truncated table Message-ID: In-Reply-To: <20260624181426.NJDNpVd1RE-qJjBVh5jtQg@awo.kakao.com> On Wed, Jun 24, 2026 at 06:14:26PM +0900, oxsignal wrote: > Hi Patrick, > > Thanks for the patch series, for adding the dedicated reftable fuzzer, and for > the credit. > > I reviewed the cover letter and the reftable hardening patches. Patch 05/11 > matches the OOB write case I reported: > the new minimum block-size validation before handling the log block prevents > the bogus inflated-size underflow from reaching the inflate/copy path. > > The rest of the series also looks like a good cleanup of the corrupted reftable > parser surface, especially the restart-count/restart-offset and truncated-table > checks. > If I find any remaining malformed-table case that is not covered by this > series, I will follow up with the reproducer. > > Thanks again for handling this so quickly. Perfect, thanks for the report and reading through the patches! Patrick