Merge patch series "netfs, ceph, nfs, cachefiles: Miscellaneous fixes/changes"

David Howells <dhowells@redhat.com> says:

Here are some miscellaneous fixes and changes for netfslib and the ceph and
nfs filesystems:

 (1) Ignore silly-rename files from afs and nfs when building the header
     archive in a kernel build.

 (2) netfs: Fix the way read result collection applies results to folios
     when each folio is being read by multiple subrequests and the results
     come out of order.

 (3) netfs: Fix ENOMEM handling in buffered reads.

 (4) nfs: Fix an oops in nfs_netfs_init_request() when copying to the cache.

 (5) cachefiles: Parse the "secctx" command immediately to get the correct
     error rather than leaving it to the "bind" command.

 (6) netfs: Remove a redundant smp_rmb().  This isn't a bug per se and
     could be deferred.

 (7) netfs: Fix missing barriers by using clear_and_wake_up_bit().

 (8) netfs: Work around recursion in read retry by failing and abandoning
     the retried subrequest if no I/O is performed.

     [!] NOTE: This only works around the recursion problem if the
     	 recursion keeps returning no data.  If the server manages, say, to
     	 repeatedly return a single byte of data faster than the retry
     	 algorithm can complete, it will still recurse and the stack
     	 overrun may still occur.  Actually fixing this requires quite an
     	 intrusive change which will hopefully make the next merge window.

 (9) netfs: Fix the clearance of a folio_queue when unlocking the page if
     we're going to want to subsequently send the queue for copying to the
     cache (if, for example, we're using ceph).

(10) netfs: Fix the lack of cancellation of copy-to-cache when the cache
     for a file is temporarily disabled (for example when a DIO write is
     done to the file).  This patch and (9) fix hangs with ceph.

With these patches, I can run xfstest -g quick to completion on ceph with a
local cache.

The patches can also be found here with a bonus cifs patch:

	https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/log/?h=netfs-fixes

* patches from https://lore.kernel.org/r/20241213135013.2964079-1-dhowells@redhat.com:
  netfs: Fix is-caching check in read-retry
  netfs: Fix the (non-)cancellation of copy when cache is temporarily disabled
  netfs: Fix ceph copy to cache on write-begin
  netfs: Work around recursion by abandoning retry if nothing read
  netfs: Fix missing barriers by using clear_and_wake_up_bit()
  netfs: Remove redundant use of smp_rmb()
  cachefiles: Parse the "secctx" immediately
  nfs: Fix oops in nfs_netfs_init_request() when copying to cache
  netfs: Fix enomem handling in buffered reads
  netfs: Fix non-contiguous donation between completed reads
  kheaders: Ignore silly-rename files

Link: https://lore.kernel.org/r/20241213135013.2964079-1-dhowells@redhat.com
Signed-off-by: Christian Brauner <brauner@kernel.org>

Author: brauner

fs/9p/vfs_addr.c

@@ -57,6 +57,8 @@ static void v9fs_issue_write(struct netfs_io_subrequest *subreq)
 	int err, len;
 
 	len = p9_client_write(fid, subreq->start, &subreq->io_iter, &err);
+	if (len > 0)
+		__set_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags);
 	netfs_write_subrequest_terminated(subreq, len ?: err, false);
 }
 
@@ -80,8 +82,10 @@ static void v9fs_issue_read(struct netfs_io_subrequest *subreq)
 	if (pos + total >= i_size_read(rreq->inode))
 		__set_bit(NETFS_SREQ_HIT_EOF, &subreq->flags);
 
-	if (!err)
+	if (!err) {
 		subreq->transferred += total;
+		__set_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags);
+	}
 
 	netfs_read_subreq_terminated(subreq, err, false);
 }

fs/afs/write.c

@@ -122,7 +122,7 @@ static void afs_issue_write_worker(struct work_struct *work)
 	if (subreq->debug_index == 3)
 		return netfs_write_subrequest_terminated(subreq, -ENOANO, false);
 
-	if (!test_bit(NETFS_SREQ_RETRYING, &subreq->flags)) {
+	if (!subreq->retry_count) {
 		set_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags);
 		return netfs_write_subrequest_terminated(subreq, -EAGAIN, false);
 	}
@@ -149,6 +149,9 @@ static void afs_issue_write_worker(struct work_struct *work)
 	afs_wait_for_operation(op);
 	ret = afs_put_operation(op);
 	switch (ret) {
+	case 0:
+		__set_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags);
+		break;
 	case -EACCES:
 	case -EPERM:
 	case -ENOKEY:

fs/cachefiles/daemon.c

@@ -15,6 +15,7 @@
 #include <linux/namei.h>
 #include <linux/poll.h>
 #include <linux/mount.h>
+#include <linux/security.h>
 #include <linux/statfs.h>
 #include <linux/ctype.h>
 #include <linux/string.h>
@@ -576,7 +577,7 @@ static int cachefiles_daemon_dir(struct cachefiles_cache *cache, char *args)
  */
 static int cachefiles_daemon_secctx(struct cachefiles_cache *cache, char *args)
 {
-	char *secctx;
+	int err;
 
 	_enter(",%s", args);
 
@@ -585,16 +586,16 @@ static int cachefiles_daemon_secctx(struct cachefiles_cache *cache, char *args)
 		return -EINVAL;
 	}
 
-	if (cache->secctx) {
+	if (cache->have_secid) {
 		pr_err("Second security context specified\n");
 		return -EINVAL;
 	}
 
-	secctx = kstrdup(args, GFP_KERNEL);
-	if (!secctx)
-		return -ENOMEM;
+	err = security_secctx_to_secid(args, strlen(args), &cache->secid);
+	if (err)
+		return err;
 
-	cache->secctx = secctx;
+	cache->have_secid = true;
 	return 0;
 }
 
@@ -820,7 +821,6 @@ static void cachefiles_daemon_unbind(struct cachefiles_cache *cache)
 	put_cred(cache->cache_cred);
 
 	kfree(cache->rootdirname);
-	kfree(cache->secctx);
 	kfree(cache->tag);
 
 	_leave("");

fs/cachefiles/internal.h

@@ -122,14 +122,15 @@ struct cachefiles_cache {
 #define CACHEFILES_STATE_CHANGED	3	/* T if state changed (poll trigger) */
 #define CACHEFILES_ONDEMAND_MODE	4	/* T if in on-demand read mode */
 	char				*rootdirname;	/* name of cache root directory */
-	char				*secctx;	/* LSM security context */
 	char				*tag;		/* cache binding tag */
 	refcount_t			unbind_pincount;/* refcount to do daemon unbind */
 	struct xarray			reqs;		/* xarray of pending on-demand requests */
 	unsigned long			req_id_next;
 	struct xarray			ondemand_ids;	/* xarray for ondemand_id allocation */
 	u32				ondemand_id_next;
 	u32				msg_id_next;
+	u32				secid;		/* LSM security id */
+	bool				have_secid;	/* whether "secid" was set */
 };
 
 static inline bool cachefiles_in_ondemand_mode(struct cachefiles_cache *cache)

fs/cachefiles/security.c

@@ -18,16 +18,16 @@ int cachefiles_get_security_ID(struct cachefiles_cache *cache)
 	struct cred *new;
 	int ret;
 
-	_enter("{%s}", cache->secctx);
+	_enter("{%u}", cache->have_secid ? cache->secid : 0);
 
 	new = prepare_kernel_cred(current);
 	if (!new) {
 		ret = -ENOMEM;
 		goto error;
 	}
 
-	if (cache->secctx) {
-		ret = set_security_override_from_ctx(new, cache->secctx);
+	if (cache->have_secid) {
+		ret = set_security_override(new, cache->secid);
 		if (ret < 0) {
 			put_cred(new);
 			pr_err("Security denies permission to nominate security context: error %d\n",

fs/netfs/buffered_read.c

@@ -275,22 +275,14 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq)
 			netfs_stat(&netfs_n_rh_download);
 			if (rreq->netfs_ops->prepare_read) {
 				ret = rreq->netfs_ops->prepare_read(subreq);
-				if (ret < 0) {
-					atomic_dec(&rreq->nr_outstanding);
-					netfs_put_subrequest(subreq, false,
-							     netfs_sreq_trace_put_cancel);
-					break;
-				}
+				if (ret < 0)
+					goto prep_failed;
 				trace_netfs_sreq(subreq, netfs_sreq_trace_prepare);
 			}
 
 			slice = netfs_prepare_read_iterator(subreq);
-			if (slice < 0) {
-				atomic_dec(&rreq->nr_outstanding);
-				netfs_put_subrequest(subreq, false, netfs_sreq_trace_put_cancel);
-				ret = slice;
-				break;
-			}
+			if (slice < 0)
+				goto prep_iter_failed;
 
 			rreq->netfs_ops->issue_read(subreq);
 			goto done;
@@ -302,6 +294,8 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq)
 			trace_netfs_sreq(subreq, netfs_sreq_trace_submit);
 			netfs_stat(&netfs_n_rh_zero);
 			slice = netfs_prepare_read_iterator(subreq);
+			if (slice < 0)
+				goto prep_iter_failed;
 			__set_bit(NETFS_SREQ_CLEAR_TAIL, &subreq->flags);
 			netfs_read_subreq_terminated(subreq, 0, false);
 			goto done;
@@ -310,6 +304,8 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq)
 		if (source == NETFS_READ_FROM_CACHE) {
 			trace_netfs_sreq(subreq, netfs_sreq_trace_submit);
 			slice = netfs_prepare_read_iterator(subreq);
+			if (slice < 0)
+				goto prep_iter_failed;
 			netfs_read_cache_to_pagecache(rreq, subreq);
 			goto done;
 		}
@@ -318,6 +314,14 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq)
 		WARN_ON_ONCE(1);
 		break;
 
+	prep_iter_failed:
+		ret = slice;
+	prep_failed:
+		subreq->error = ret;
+		atomic_dec(&rreq->nr_outstanding);
+		netfs_put_subrequest(subreq, false, netfs_sreq_trace_put_cancel);
+		break;
+
 	done:
 		size -= slice;
 		start += slice;

fs/netfs/direct_write.c

@@ -104,7 +104,6 @@ ssize_t netfs_unbuffered_write_iter_locked(struct kiocb *iocb, struct iov_iter *
 		trace_netfs_rreq(wreq, netfs_rreq_trace_wait_ip);
 		wait_on_bit(&wreq->flags, NETFS_RREQ_IN_PROGRESS,
 			    TASK_UNINTERRUPTIBLE);
-		smp_rmb(); /* Read error/transferred after RIP flag */
 		ret = wreq->error;
 		if (ret == 0) {
 			ret = wreq->transferred;

fs/netfs/read_collect.c

@@ -62,10 +62,14 @@ static void netfs_unlock_read_folio(struct netfs_io_subrequest *subreq,
 		} else {
 			trace_netfs_folio(folio, netfs_folio_trace_read_done);
 		}
+
+		folioq_clear(folioq, slot);
 	} else {
 		// TODO: Use of PG_private_2 is deprecated.
 		if (test_bit(NETFS_SREQ_COPY_TO_CACHE, &subreq->flags))
 			netfs_pgpriv2_mark_copy_to_cache(subreq, rreq, folioq, slot);
+		else
+			folioq_clear(folioq, slot);
 	}
 
 	if (!test_bit(NETFS_RREQ_DONT_UNLOCK_FOLIOS, &rreq->flags)) {
@@ -77,8 +81,6 @@ static void netfs_unlock_read_folio(struct netfs_io_subrequest *subreq,
 			folio_unlock(folio);
 		}
 	}
-
-	folioq_clear(folioq, slot);
 }
 
 /*
@@ -247,16 +249,17 @@ static bool netfs_consume_read_data(struct netfs_io_subrequest *subreq, bool was
 
 	/* Deal with the trickiest case: that this subreq is in the middle of a
 	 * folio, not touching either edge, but finishes first.  In such a
-	 * case, we donate to the previous subreq, if there is one, so that the
-	 * donation is only handled when that completes - and remove this
-	 * subreq from the list.
+	 * case, we donate to the previous subreq, if there is one and if it is
+	 * contiguous, so that the donation is only handled when that completes
+	 * - and remove this subreq from the list.
 	 *
 	 * If the previous subreq finished first, we will have acquired their
 	 * donation and should be able to unlock folios and/or donate nextwards.
 	 */
 	if (!subreq->consumed &&
 	    !prev_donated &&
-	    !list_is_first(&subreq->rreq_link, &rreq->subrequests)) {
+	    !list_is_first(&subreq->rreq_link, &rreq->subrequests) &&
+	    subreq->start == prev->start + prev->len) {
 		prev = list_prev_entry(subreq, rreq_link);
 		WRITE_ONCE(prev->next_donated, prev->next_donated + subreq->len);
 		subreq->start += subreq->len;
@@ -378,8 +381,7 @@ static void netfs_rreq_assess(struct netfs_io_request *rreq)
 	task_io_account_read(rreq->transferred);
 
 	trace_netfs_rreq(rreq, netfs_rreq_trace_wake_ip);
-	clear_bit_unlock(NETFS_RREQ_IN_PROGRESS, &rreq->flags);
-	wake_up_bit(&rreq->flags, NETFS_RREQ_IN_PROGRESS);
+	clear_and_wake_up_bit(NETFS_RREQ_IN_PROGRESS, &rreq->flags);
 
 	trace_netfs_rreq(rreq, netfs_rreq_trace_done);
 	netfs_clear_subrequests(rreq, false);
@@ -438,7 +440,7 @@ void netfs_read_subreq_progress(struct netfs_io_subrequest *subreq,
 	     rreq->origin == NETFS_READPAGE ||
 	     rreq->origin == NETFS_READ_FOR_WRITE)) {
 		netfs_consume_read_data(subreq, was_async);
-		__clear_bit(NETFS_SREQ_NO_PROGRESS, &subreq->flags);
+		__set_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags);
 	}
 }
 EXPORT_SYMBOL(netfs_read_subreq_progress);
@@ -497,7 +499,7 @@ void netfs_read_subreq_terminated(struct netfs_io_subrequest *subreq,
 		     rreq->origin == NETFS_READPAGE ||
 		     rreq->origin == NETFS_READ_FOR_WRITE)) {
 			netfs_consume_read_data(subreq, was_async);
-			__clear_bit(NETFS_SREQ_NO_PROGRESS, &subreq->flags);
+			__set_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags);
 		}
 		rreq->transferred += subreq->transferred;
 	}
@@ -511,10 +513,13 @@ void netfs_read_subreq_terminated(struct netfs_io_subrequest *subreq,
 		} else {
 			trace_netfs_sreq(subreq, netfs_sreq_trace_short);
 			if (subreq->transferred > subreq->consumed) {
-				__set_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags);
-				__clear_bit(NETFS_SREQ_NO_PROGRESS, &subreq->flags);
-				set_bit(NETFS_RREQ_NEED_RETRY, &rreq->flags);
-			} else if (!__test_and_set_bit(NETFS_SREQ_NO_PROGRESS, &subreq->flags)) {
+				/* If we didn't read new data, abandon retry. */
+				if (subreq->retry_count &&
+				    test_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags)) {
+					__set_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags);
+					set_bit(NETFS_RREQ_NEED_RETRY, &rreq->flags);
+				}
+			} else if (test_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags)) {
 				__set_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags);
 				set_bit(NETFS_RREQ_NEED_RETRY, &rreq->flags);
 			} else {

fs/netfs/read_pgpriv2.c

@@ -170,6 +170,10 @@ void netfs_pgpriv2_write_to_the_cache(struct netfs_io_request *rreq)
 
 	trace_netfs_write(wreq, netfs_write_trace_copy_to_cache);
 	netfs_stat(&netfs_n_wh_copy_to_cache);
+	if (!wreq->io_streams[1].avail) {
+		netfs_put_request(wreq, false, netfs_rreq_trace_put_return);
+		goto couldnt_start;
+	}
 
 	for (;;) {
 		error = netfs_pgpriv2_copy_folio(wreq, folio);

fs/netfs/read_retry.c

@@ -49,13 +49,15 @@ static void netfs_retry_read_subrequests(struct netfs_io_request *rreq)
 	 * up to the first permanently failed one.
 	 */
 	if (!rreq->netfs_ops->prepare_read &&
-	    !test_bit(NETFS_RREQ_COPY_TO_CACHE, &rreq->flags)) {
+	    !rreq->cache_resources.ops) {
 		struct netfs_io_subrequest *subreq;
 
 		list_for_each_entry(subreq, &rreq->subrequests, rreq_link) {
 			if (test_bit(NETFS_SREQ_FAILED, &subreq->flags))
 				break;
 			if (__test_and_clear_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags)) {
+				__clear_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags);
+				subreq->retry_count++;
 				netfs_reset_iter(subreq);
 				netfs_reissue_read(rreq, subreq);
 			}
@@ -137,7 +139,8 @@ static void netfs_retry_read_subrequests(struct netfs_io_request *rreq)
 			stream0->sreq_max_len = subreq->len;
 
 			__clear_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags);
-			__set_bit(NETFS_SREQ_RETRYING, &subreq->flags);
+			__clear_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags);
+			subreq->retry_count++;
 
 			spin_lock_bh(&rreq->lock);
 			list_add_tail(&subreq->rreq_link, &rreq->subrequests);
@@ -213,7 +216,6 @@ static void netfs_retry_read_subrequests(struct netfs_io_request *rreq)
 			subreq->error = -ENOMEM;
 		__clear_bit(NETFS_SREQ_FAILED, &subreq->flags);
 		__clear_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags);
-		__clear_bit(NETFS_SREQ_RETRYING, &subreq->flags);
 	}
 	spin_lock_bh(&rreq->lock);
 	list_splice_tail_init(&queue, &rreq->subrequests);