virtio_blk: set the zone write granularitygrafted

virtblk_read_zoned_limits() reads the write granularity that the device
reports in virtio_blk_zoned_characteristics and assigns it to the
physical block size and to io_min, but never to the limit that is named
after it. queue_limits.zone_write_granularity is left at zero, so
blk_validate_zoned_limits() raises it to the logical block size:

	if (lim->zone_write_granularity < lim->logical_block_size)
		lim->zone_write_granularity = lim->logical_block_size;

A device that reports a granularity coarser than its logical block size,
which is what the field exists to express, therefore has it silently
reduced. A 512e host managed disk passed through to a guest reports a
logical block size of 512 and a write granularity of 4096, and the guest
ends up with a zone write granularity of 512.

bio_split_alignment() returns lim->zone_write_granularity if it is non-zero
and bio_split_io_at() may split a bio with as per bio_split_alignment().
This can real to the write getting rejected by the host drive, as the write
is not aligned to the physical block size.

zonefs also takes its block size from bdev_zone_write_granularity(), so it
would incorrectly use 512 on a disk that requires 4096.

sd_zbc_read_zones() sets the limit from the physical block size for the
same reason. NVMe ZNS and null_blk leave it unset, but the fallback
gives the right answer for them, as their write granularity is the
logical block size. virtio carries a separate value that may exceed it.

Set the zone write granularity from the value that the device reports.

Fixes: 95bfec41bd3d ("virtio-blk: add support for zoned block devices")
Signed-off-by: Niklas Cassel <cassel@kernel.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://patch.msgid.link/20260918140641.2031075-2-cassel@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 149 insertions, 0 deletions

diff --git a/fs/dlm/requestqueue.c b/fs/dlm/requestqueue.c
new file mode 100644
index 000000000..719a5243a
--- /dev/null
+++ b/fs/dlm/requestqueue.c
@@ -0,0 +1,149 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/******************************************************************************
+*******************************************************************************
+**
+**  Copyright (C) 2005-2007 Red Hat, Inc.  All rights reserved.
+**
+**
+*******************************************************************************
+******************************************************************************/
+
+#include "dlm_internal.h"
+#include "member.h"
+#include "lock.h"
+#include "dir.h"
+#include "config.h"
+#include "requestqueue.h"
+#include "util.h"
+
+struct rq_entry {
+	struct list_head list;
+	uint32_t recover_seq;
+	int nodeid;
+	struct dlm_message request;
+};
+
+/*
+ * Requests received while the lockspace is in recovery get added to the
+ * request queue and processed when recovery is complete.  This happens when
+ * the lockspace is suspended on some nodes before it is on others, or the
+ * lockspace is enabled on some while still suspended on others.
+ */
+
+void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid,
+			  const struct dlm_message *ms)
+{
+	struct rq_entry *e;
+	int length = le16_to_cpu(ms->m_header.h_length) -
+		sizeof(struct dlm_message);
+
+	e = kmalloc(sizeof(struct rq_entry) + length, GFP_ATOMIC);
+	if (!e) {
+		log_print("dlm_add_requestqueue: out of memory len %d", length);
+		return;
+	}
+
+	e->recover_seq = ls->ls_recover_seq & 0xFFFFFFFF;
+	e->nodeid = nodeid;
+	memcpy(&e->request, ms, sizeof(*ms));
+	memcpy(&e->request.m_extra, ms->m_extra, length);
+
+	list_add_tail(&e->list, &ls->ls_requestqueue);
+}
+
+/*
+ * Called by dlm_recoverd to process normal messages saved while recovery was
+ * happening.  Normal locking has been enabled before this is called.  dlm_recv
+ * upon receiving a message, will wait for all saved messages to be drained
+ * here before processing the message it got.  If a new dlm_ls_stop() arrives
+ * while we're processing these saved messages, it may block trying to suspend
+ * dlm_recv if dlm_recv is waiting for us in dlm_wait_requestqueue.  In that
+ * case, we don't abort since locking_stopped is still 0.  If dlm_recv is not
+ * waiting for us, then this processing may be aborted due to locking_stopped.
+ */
+
+int dlm_process_requestqueue(struct dlm_ls *ls)
+{
+	struct rq_entry *e;
+	struct dlm_message *ms;
+	int error = 0;
+
+	write_lock_bh(&ls->ls_requestqueue_lock);
+	for (;;) {
+		if (list_empty(&ls->ls_requestqueue)) {
+			clear_bit(LSFL_RECV_MSG_BLOCKED, &ls->ls_flags);
+			error = 0;
+			break;
+		}
+		e = list_first_entry(&ls->ls_requestqueue, struct rq_entry, list);
+
+		ms = &e->request;
+
+		log_limit(ls, "dlm_process_requestqueue msg %d from %d "
+			  "lkid %x remid %x result %d seq %u",
+			  le32_to_cpu(ms->m_type),
+			  le32_to_cpu(ms->m_header.h_nodeid),
+			  le32_to_cpu(ms->m_lkid), le32_to_cpu(ms->m_remid),
+			  from_dlm_errno(le32_to_cpu(ms->m_result)),
+			  e->recover_seq);
+
+		dlm_receive_message_saved(ls, &e->request, e->recover_seq);
+		list_del(&e->list);
+		kfree(e);
+
+		if (dlm_locking_stopped(ls)) {
+			log_debug(ls, "process_requestqueue abort running");
+			error = -EINTR;
+			break;
+		}
+		write_unlock_bh(&ls->ls_requestqueue_lock);
+		schedule();
+		write_lock_bh(&ls->ls_requestqueue_lock);
+	}
+	write_unlock_bh(&ls->ls_requestqueue_lock);
+
+	return error;
+}
+
+static int purge_request(struct dlm_ls *ls, struct dlm_message *ms, int nodeid)
+{
+	__le32 type = ms->m_type;
+
+	/* the ls is being cleaned up and freed by release_lockspace */
+	if (!atomic_read(&ls->ls_count))
+		return 1;
+
+	if (dlm_is_removed(ls, nodeid))
+		return 1;
+
+	/* directory operations are always purged because the directory is
+	   always rebuilt during recovery and the lookups resent */
+
+	if (type == cpu_to_le32(DLM_MSG_REMOVE) ||
+	    type == cpu_to_le32(DLM_MSG_LOOKUP) ||
+	    type == cpu_to_le32(DLM_MSG_LOOKUP_REPLY))
+		return 1;
+
+	if (!dlm_no_directory(ls))
+		return 0;
+
+	return 1;
+}
+
+void dlm_purge_requestqueue(struct dlm_ls *ls)
+{
+	struct dlm_message *ms;
+	struct rq_entry *e, *safe;
+
+	write_lock_bh(&ls->ls_requestqueue_lock);
+	list_for_each_entry_safe(e, safe, &ls->ls_requestqueue, list) {
+		ms =  &e->request;
+
+		if (purge_request(ls, ms, e->nodeid)) {
+			list_del(&e->list);
+			kfree(e);
+		}
+	}
+	write_unlock_bh(&ls->ls_requestqueue_lock);
+}
+