On Mon, Sep 27, 2021 at 11:22 PM Eric Biggers <ebiggers@xxxxxxxxxx> wrote:
>
> Fix an integer overflow when computing the Merkle tree layout of
> extremely large files, exposed by btrfs adding support for fs-verity.
I wonder if 'i_size' should be u64. I'm not convinced people think
about 'loff_t' being signed - but while that's required for negative
lseek() offsets, I'm not sure it makes tons of sense for an inode
size.
Same goes for f_pos, for that matter.
But who knows what games people have played with magic numbers (ie
"-1") internally, or where they _want_ signed compares. So it's
certainly not some obvious trivial fix.
Pulled.
Linus
|