On Tue, 2017-02-28 at 12:12 +0200, Boaz Harrosh wrote: > On 02/28/2017 03:11 AM, Jeff Layton wrote: > <> > > > > I'll probably have questions about the read side as well, but for now it > > looks like it's mostly used in an ad-hoc way to communicate errors > > across subsystems (block to fs layer, for instance). > > If memory does not fail me it used to be checked long time ago in the > read-ahead case. On the buffered read case, the first page is read synchronous > and any error is returned to the caller, but then a read-ahead chunk is > read async all the while the original thread returned to the application. > So any errors are only recorded on the page-bit, since otherwise the uptodate > is off and the IO will be retransmitted. Then the move to read_iter changed > all that I think. > But again this is like 5-6 years ago, and maybe I didn't even understand > very well. > Yep, that's what I meant about using it to communicate errors between layers. e.g. end_buffer_async_read will check PageError and only SetPageUptodate if it's not set. That has morphed a lot in the last few years though and it looks like it may rely on PG_error less than it used to. > > I would like a Documentation of all this as well please. Where are the > tests for this? > Documentation is certainly doable (and I'd like to write some once we have this all straightened out). In particular, I think we need clear guidelines for fs authors on how to handle pagecache read and write errors. Tests are a little tougher -- this is all kernel-internal stuff and not easily visible to userland. The one thing I have noticed is that even if you set AS_ENOSPC in the mapping, you'll still get back -EIO on the first fsync if any PG_error bits are set. I think we ought to fix that by not doing the TestClearPageError call in __filemap_fdatawait_range, and just rely on the mapping error there. We could maybe roll a test for that, but it's rather hard to test ENOSPC conditions in a fs-agnostic way. I'm open to suggestions here though. -- Jeff Layton <jlayton@xxxxxxxxxx>
