netfs: Fix missing xas_retry() calls in xarray iteration

To be applied to linux-5.15.y.

commit 59d0d52c30d4991ac4b329f049cc37118e00f5b0 upstream

Stops kernel from crashing when encountering an XAS retry entry. Patch modified from upstream to work with pages instead of folios. Also omits fixes to "dodgy maths" as unrelated to fixing the crash.

---
 fs/netfs/read_helper.c | 6 ++++++
 1 file changed, 6 insertions(+)

index 242f8bcb34a4..4de15555bceb 100644
@@ -248,6 +248,9 @@ static void netfs_rreq_unmark_after_write(struct netfs_read_request *rreq,
 		XA_STATE(xas, &rreq->mapping->i_pages, subreq->start / PAGE_SIZE);
 
 		xas_for_each(&xas, page, (subreq->start + subreq->len - 1) / PAGE_SIZE) {
+			if(xas_retry(&xas, page))
+				continue;
+
 			/* We might have multiple writes from the same huge
 			 * page, but we mustn't unlock a page more than once.
 			 */

@@ -403,6 +406,9 @@ static void netfs_rreq_unlock(struct netfs_read_request *rreq)
 		unsigned int pgend = pgpos + thp_size(page);
 		bool pg_failed = false;
 
+		if(xas_retry(&xas, page))
+			continue;
+
 		for (;;) {
 			if (!subreq) {
 				pg_failed = true;

2.30.2

That is not this commit at all :(

Thank you for your time.

From what I can tell, the upstream commit merged by Linus includes two unrelated sets of changes: handling xas_retry, and fixes to "dodgy maths". I discarded the fixes to dodgy maths for two reasons, first the commit log says they solve a theoretical potential problem and the guidelines for submissions to the stable kernel say to avoid such patches, and second, I lack the expertise to be confident those fixes are correct when working with pages and not folios.

Unsure what I should have done here.

Submit a real patch for the normal tree and then we can backport the same commit id to a stable tree. You can't backport a merge commit as obviously that doesn't make sense in an older tree.

thanks,

greg k-h

No, this is a merge commit, sorry. To quote my bot:

<formletter>

This is not the correct way to submit patches for inclusion in the stable kernel tree. Please read:
https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
for how to do this properly.

</formletter>