Re: bvec_iter.bi_sector -> loff_t?
From: Christoph Hellwig <hch@lst.de>
Date: 2024-06-20 15:20:31
Also in:
linux-bcachefs, linux-fsdevel
From: Christoph Hellwig <hch@lst.de>
Date: 2024-06-20 15:20:31
Also in:
linux-bcachefs, linux-fsdevel
On Thu, Jun 20, 2024 at 04:15:23PM +0100, Matthew Wilcox wrote:
quoted
and have O_DIRECT with a 32-bit memory alignment work just fine, where before it would EINVAL. The sector size memory alignment thing has always been odd and never rooted in anything other than "oh let's just require the whole combination of size/disk offset/alignment to be sector based".Oh, cool! https://man7.org/linux/man-pages/man2/open.2.html doesn't know about this yet; is anyone working on updating it?
Just remember that there are two kinds of alignments: - the memory alignment, which Jens is talking about - the offset/size alignment, which is set by the LBA size statx (optionally) exposes both in the stx_dio_mem_align and stx_dio_offset_align fields, which are documented in the statx(2) man page. For network file systems like nfs there might be no alignment requirements at all.