Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

Cross-merge networking fixes after downstream PR.

Conflicts:

drivers/net/ethernet/broadcom/bnxt/bnxt.c
  f7ce5eb2cb79 ("bnxt_en: Fix crash in bnxt_get_max_rss_ctx_ring()")
  20c8ad72eb7f ("eth: bnxt: use the RSS context XArray instead of the local list")

Adjacent changes:

net/ethtool/ioctl.c
  503757c80928 ("net: ethtool: Fix RSS setting")
  eac9122f0c41 ("net: ethtool: record custom RSS contexts in the XArray")

Signed-off-by: Jakub Kicinski <kuba@kernel.org>

14 files changed, 184 insertions, 53 deletions

diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c
index 7c72a9e6f511..a0deb8266011 100644
--- a/fs/bcachefs/btree_gc.c
+++ b/fs/bcachefs/btree_gc.c
@@ -641,16 +641,30 @@ static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree, bool in
 		target_depth = 0;
 
 	/* root */
-	mutex_lock(&c->btree_root_lock);
-	struct btree *b = bch2_btree_id_root(c, btree)->b;
-	if (!btree_node_fake(b)) {
+	do {
+retry_root:
+		bch2_trans_begin(trans);
+
+		struct btree_iter iter;
+		bch2_trans_node_iter_init(trans, &iter, btree, POS_MIN,
+					  0, bch2_btree_id_root(c, btree)->b->c.level, 0);
+		struct btree *b = bch2_btree_iter_peek_node(&iter);
+		ret = PTR_ERR_OR_ZERO(b);
+		if (ret)
+			goto err_root;
+
+		if (b != btree_node_root(c, b)) {
+			bch2_trans_iter_exit(trans, &iter);
+			goto retry_root;
+		}
+
 		gc_pos_set(c, gc_pos_btree(btree, b->c.level + 1, SPOS_MAX));
-		ret = lockrestart_do(trans,
-			bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1,
-					 NULL, NULL, bkey_i_to_s_c(&b->key), initial));
+		struct bkey_s_c k = bkey_i_to_s_c(&b->key);
+		ret = bch2_gc_mark_key(trans, btree, b->c.level + 1, NULL, NULL, k, initial);
 		level = b->c.level;
-	}
-	mutex_unlock(&c->btree_root_lock);
+err_root:
+		bch2_trans_iter_exit(trans, &iter);
+	} while (bch2_err_matches(ret, BCH_ERR_transaction_restart));
 
 	if (ret)
 		return ret;
diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c
index 0ed9e6574fcd..19352a08ea20 100644
--- a/fs/bcachefs/btree_iter.c
+++ b/fs/bcachefs/btree_iter.c
@@ -996,7 +996,7 @@ retry_all:
 
 	bch2_trans_unlock(trans);
 	cond_resched();
-	trans->locked = true;
+	trans_set_locked(trans);
 
 	if (unlikely(trans->memory_allocation_failure)) {
 		struct closure cl;
@@ -3089,7 +3089,8 @@ u32 bch2_trans_begin(struct btree_trans *trans)
 		bch2_trans_srcu_unlock(trans);
 
 	trans->last_begin_ip = _RET_IP_;
-	trans->locked  = true;
+
+	trans_set_locked(trans);
 
 	if (trans->restarted) {
 		bch2_btree_path_traverse_all(trans);
@@ -3159,7 +3160,6 @@ got_trans:
 	trans->last_begin_time	= local_clock();
 	trans->fn_idx		= fn_idx;
 	trans->locking_wait.task = current;
-	trans->locked		= true;
 	trans->journal_replay_not_finished =
 		unlikely(!test_bit(JOURNAL_replay_done, &c->journal.flags)) &&
 		atomic_inc_not_zero(&c->journal_keys.ref);
@@ -3193,6 +3193,7 @@ got_trans:
 	trans->srcu_idx		= srcu_read_lock(&c->btree_trans_barrier);
 	trans->srcu_lock_time	= jiffies;
 	trans->srcu_held	= true;
+	trans_set_locked(trans);
 
 	closure_init_stack_release(&trans->ref);
 	return trans;
diff --git a/fs/bcachefs/btree_locking.c b/fs/bcachefs/btree_locking.c
index d66fff22109a..c51826fd557f 100644
--- a/fs/bcachefs/btree_locking.c
+++ b/fs/bcachefs/btree_locking.c
@@ -231,7 +231,7 @@ static noinline int break_cycle(struct lock_graph *g, struct printbuf *cycle)
 			prt_newline(&buf);
 		}
 
-		bch2_print_string_as_lines(KERN_ERR, buf.buf);
+		bch2_print_string_as_lines_nonblocking(KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
 		BUG();
 	}
@@ -792,7 +792,7 @@ static inline int __bch2_trans_relock(struct btree_trans *trans, bool trace)
 			return bch2_trans_relock_fail(trans, path, &f, trace);
 	}
 
-	trans->locked = true;
+	trans_set_locked(trans);
 out:
 	bch2_trans_verify_locks(trans);
 	return 0;
@@ -812,16 +812,14 @@ void bch2_trans_unlock_noassert(struct btree_trans *trans)
 {
 	__bch2_trans_unlock(trans);
 
-	trans->locked = false;
-	trans->last_unlock_ip = _RET_IP_;
+	trans_set_unlocked(trans);
 }
 
 void bch2_trans_unlock(struct btree_trans *trans)
 {
 	__bch2_trans_unlock(trans);
 
-	trans->locked = false;
-	trans->last_unlock_ip = _RET_IP_;
+	trans_set_unlocked(trans);
 }
 
 void bch2_trans_unlock_long(struct btree_trans *trans)
diff --git a/fs/bcachefs/btree_locking.h b/fs/bcachefs/btree_locking.h
index 7f41545b9147..75a6274c7d27 100644
--- a/fs/bcachefs/btree_locking.h
+++ b/fs/bcachefs/btree_locking.h
@@ -193,6 +193,28 @@ int bch2_six_check_for_deadlock(struct six_lock *lock, void *p);
 
 /* lock: */
 
+static inline void trans_set_locked(struct btree_trans *trans)
+{
+	if (!trans->locked) {
+		trans->locked = true;
+		trans->last_unlock_ip = 0;
+
+		trans->pf_memalloc_nofs = (current->flags & PF_MEMALLOC_NOFS) != 0;
+		current->flags |= PF_MEMALLOC_NOFS;
+	}
+}
+
+static inline void trans_set_unlocked(struct btree_trans *trans)
+{
+	if (trans->locked) {
+		trans->locked = false;
+		trans->last_unlock_ip = _RET_IP_;
+
+		if (!trans->pf_memalloc_nofs)
+			current->flags &= ~PF_MEMALLOC_NOFS;
+	}
+}
+
 static inline int __btree_node_lock_nopath(struct btree_trans *trans,
 					 struct btree_bkey_cached_common *b,
 					 enum six_lock_type type,
diff --git a/fs/bcachefs/btree_types.h b/fs/bcachefs/btree_types.h
index 87f485e9c552..48cb1a7d31c5 100644
--- a/fs/bcachefs/btree_types.h
+++ b/fs/bcachefs/btree_types.h
@@ -484,6 +484,7 @@ struct btree_trans {
 	bool			lock_may_not_fail:1;
 	bool			srcu_held:1;
 	bool			locked:1;
+	bool			pf_memalloc_nofs:1;
 	bool			write_locked:1;
 	bool			used_mempool:1;
 	bool			in_traverse_all:1;
diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c
index 743d57eba760..314ee3e0187f 100644
--- a/fs/bcachefs/buckets.c
+++ b/fs/bcachefs/buckets.c
@@ -805,7 +805,7 @@ int bch2_bucket_ref_update(struct btree_trans *trans, struct bch_dev *ca,
 			"bucket %u:%zu gen %u (mem gen %u) data type %s: stale dirty ptr (gen %u)\n"
 			"while marking %s",
 			ptr->dev, bucket_nr, b_gen,
-			*bucket_gen(ca, bucket_nr),
+			bucket_gen_get(ca, bucket_nr),
 			bch2_data_type_str(bucket_data_type ?: ptr_data_type),
 			ptr->gen,
 			(printbuf_reset(&buf),
diff --git a/fs/bcachefs/buckets.h b/fs/bcachefs/buckets.h
index 80ee0be9793e..8ad4be73860c 100644
--- a/fs/bcachefs/buckets.h
+++ b/fs/bcachefs/buckets.h
@@ -116,6 +116,14 @@ static inline u8 *bucket_gen(struct bch_dev *ca, size_t b)
 	return gens->b + b;
 }
 
+static inline u8 bucket_gen_get(struct bch_dev *ca, size_t b)
+{
+	rcu_read_lock();
+	u8 gen = *bucket_gen(ca, b);
+	rcu_read_unlock();
+	return gen;
+}
+
 static inline size_t PTR_BUCKET_NR(const struct bch_dev *ca,
 				   const struct bch_extent_ptr *ptr)
 {
diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index 74a1e12a7a14..fa1fee05cf8f 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -2073,8 +2073,7 @@ int __init bch2_vfs_init(void)
 {
 	int ret = -ENOMEM;
 
-	bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT |
-				      SLAB_ACCOUNT);
+	bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT);
 	if (!bch2_inode_cache)
 		goto err;
 
diff --git a/fs/bcachefs/io_misc.c b/fs/bcachefs/io_misc.c
index 4ec979b4b23e..4583c9386e8c 100644
--- a/fs/bcachefs/io_misc.c
+++ b/fs/bcachefs/io_misc.c
@@ -125,7 +125,7 @@ err_noprint:
 	bch2_bkey_buf_exit(&old, c);
 
 	if (closure_nr_remaining(&cl) != 1) {
-		bch2_trans_unlock(trans);
+		bch2_trans_unlock_long(trans);
 		closure_sync(&cl);
 	}
 
diff --git a/fs/bcachefs/util.c b/fs/bcachefs/util.c
index de331dec2a99..4ec7e44d6e36 100644
--- a/fs/bcachefs/util.c
+++ b/fs/bcachefs/util.c
@@ -252,8 +252,10 @@ void bch2_prt_u64_base2(struct printbuf *out, u64 v)
 	bch2_prt_u64_base2_nbits(out, v, fls64(v) ?: 1);
 }
 
-void bch2_print_string_as_lines(const char *prefix, const char *lines)
+static void __bch2_print_string_as_lines(const char *prefix, const char *lines,
+					 bool nonblocking)
 {
+	bool locked = false;
 	const char *p;
 
 	if (!lines) {
@@ -261,7 +263,13 @@ void bch2_print_string_as_lines(const char *prefix, const char *lines)
 		return;
 	}
 
-	console_lock();
+	if (!nonblocking) {
+		console_lock();
+		locked = true;
+	} else {
+		locked = console_trylock();
+	}
+
 	while (1) {
 		p = strchrnul(lines, '\n');
 		printk("%s%.*s\n", prefix, (int) (p - lines), lines);
@@ -269,7 +277,18 @@ void bch2_print_string_as_lines(const char *prefix, const char *lines)
 			break;
 		lines = p + 1;
 	}
-	console_unlock();
+	if (locked)
+		console_unlock();
+}
+
+void bch2_print_string_as_lines(const char *prefix, const char *lines)
+{
+	return __bch2_print_string_as_lines(prefix, lines, false);
+}
+
+void bch2_print_string_as_lines_nonblocking(const char *prefix, const char *lines)
+{
+	return __bch2_print_string_as_lines(prefix, lines, true);
 }
 
 int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *task, unsigned skipnr,
diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
index 5d2c470a49ac..5b0533ec4c7e 100644
--- a/fs/bcachefs/util.h
+++ b/fs/bcachefs/util.h
@@ -315,6 +315,7 @@ void bch2_prt_u64_base2_nbits(struct printbuf *, u64, unsigned);
 void bch2_prt_u64_base2(struct printbuf *, u64);
 
 void bch2_print_string_as_lines(const char *prefix, const char *lines);
+void bch2_print_string_as_lines_nonblocking(const char *prefix, const char *lines);
 
 typedef DARRAY(unsigned long) bch_stacktrace;
 int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *, unsigned, gfp_t);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 38cdb8875e8e..cabb558dbdaa 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -2856,6 +2856,8 @@ static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block
 	if (ret)
 		return ret;
 
+	spin_lock_init(&fs_info->extent_map_shrinker_lock);
+
 	ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
 	if (ret)
 		return ret;
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index 744e8952abb0..b4c9a6aa118c 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -1028,7 +1028,14 @@ out_free_pre:
 	return ret;
 }
 
-static long btrfs_scan_inode(struct btrfs_inode *inode, long *scanned, long nr_to_scan)
+struct btrfs_em_shrink_ctx {
+	long nr_to_scan;
+	long scanned;
+	u64 last_ino;
+	u64 last_root;
+};
+
+static long btrfs_scan_inode(struct btrfs_inode *inode, struct btrfs_em_shrink_ctx *ctx)
 {
 	const u64 cur_fs_gen = btrfs_get_fs_generation(inode->root->fs_info);
 	struct extent_map_tree *tree = &inode->extent_tree;
@@ -1057,14 +1064,25 @@ static long btrfs_scan_inode(struct btrfs_inode *inode, long *scanned, long nr_t
 	if (!down_read_trylock(&inode->i_mmap_lock))
 		return 0;
 
-	write_lock(&tree->lock);
+	/*
+	 * We want to be fast because we can be called from any path trying to
+	 * allocate memory, so if the lock is busy we don't want to spend time
+	 * waiting for it - either some task is about to do IO for the inode or
+	 * we may have another task shrinking extent maps, here in this code, so
+	 * skip this inode.
+	 */
+	if (!write_trylock(&tree->lock)) {
+		up_read(&inode->i_mmap_lock);
+		return 0;
+	}
+
 	node = rb_first_cached(&tree->map);
 	while (node) {
 		struct extent_map *em;
 
 		em = rb_entry(node, struct extent_map, rb_node);
 		node = rb_next(node);
-		(*scanned)++;
+		ctx->scanned++;
 
 		if (em->flags & EXTENT_FLAG_PINNED)
 			goto next;
@@ -1085,16 +1103,18 @@ static long btrfs_scan_inode(struct btrfs_inode *inode, long *scanned, long nr_t
 		free_extent_map(em);
 		nr_dropped++;
 next:
-		if (*scanned >= nr_to_scan)
+		if (ctx->scanned >= ctx->nr_to_scan)
 			break;
 
 		/*
-		 * Restart if we had to reschedule, and any extent maps that were
-		 * pinned before may have become unpinned after we released the
-		 * lock and took it again.
+		 * Stop if we need to reschedule or there's contention on the
+		 * lock. This is to avoid slowing other tasks trying to take the
+		 * lock and because the shrinker might be called during a memory
+		 * allocation path and we want to avoid taking a very long time
+		 * and slowing down all sorts of tasks.
 		 */
-		if (cond_resched_rwlock_write(&tree->lock))
-			node = rb_first_cached(&tree->map);
+		if (need_resched() || rwlock_needbreak(&tree->lock))
+			break;
 	}
 	write_unlock(&tree->lock);
 	up_read(&inode->i_mmap_lock);
@@ -1102,25 +1122,30 @@ next:
 	return nr_dropped;
 }
 
-static long btrfs_scan_root(struct btrfs_root *root, long *scanned, long nr_to_scan)
+static long btrfs_scan_root(struct btrfs_root *root, struct btrfs_em_shrink_ctx *ctx)
 {
-	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct btrfs_inode *inode;
 	long nr_dropped = 0;
-	u64 min_ino = fs_info->extent_map_shrinker_last_ino + 1;
+	u64 min_ino = ctx->last_ino + 1;
 
 	inode = btrfs_find_first_inode(root, min_ino);
 	while (inode) {
-		nr_dropped += btrfs_scan_inode(inode, scanned, nr_to_scan);
+		nr_dropped += btrfs_scan_inode(inode, ctx);
 
 		min_ino = btrfs_ino(inode) + 1;
-		fs_info->extent_map_shrinker_last_ino = btrfs_ino(inode);
-		iput(&inode->vfs_inode);
+		ctx->last_ino = btrfs_ino(inode);
+		btrfs_add_delayed_iput(inode);
 
-		if (*scanned >= nr_to_scan)
+		if (ctx->scanned >= ctx->nr_to_scan)
+			break;
+
+		/*
+		 * We may be called from memory allocation paths, so we don't
+		 * want to take too much time and slowdown tasks.
+		 */
+		if (need_resched())
 			break;
 
-		cond_resched();
 		inode = btrfs_find_first_inode(root, min_ino);
 	}
 
@@ -1132,14 +1157,14 @@ static long btrfs_scan_root(struct btrfs_root *root, long *scanned, long nr_to_s
 		 * inode if there is one or we will find out this was the last
 		 * one and move to the next root.
 		 */
-		fs_info->extent_map_shrinker_last_root = btrfs_root_id(root);
+		ctx->last_root = btrfs_root_id(root);
 	} else {
 		/*
 		 * No more inodes in this root, set extent_map_shrinker_last_ino to 0 so
 		 * that when processing the next root we start from its first inode.
 		 */
-		fs_info->extent_map_shrinker_last_ino = 0;
-		fs_info->extent_map_shrinker_last_root = btrfs_root_id(root) + 1;
+		ctx->last_ino = 0;
+		ctx->last_root = btrfs_root_id(root) + 1;
 	}
 
 	return nr_dropped;
@@ -1147,19 +1172,41 @@ static long btrfs_scan_root(struct btrfs_root *root, long *scanned, long nr_to_s
 
 long btrfs_free_extent_maps(struct btrfs_fs_info *fs_info, long nr_to_scan)
 {
-	const u64 start_root_id = fs_info->extent_map_shrinker_last_root;
-	u64 next_root_id = start_root_id;
+	struct btrfs_em_shrink_ctx ctx;
+	u64 start_root_id;
+	u64 next_root_id;
 	bool cycled = false;
 	long nr_dropped = 0;
-	long scanned = 0;
+
+	ctx.scanned = 0;
+	ctx.nr_to_scan = nr_to_scan;
+
+	/*
+	 * In case we have multiple tasks running this shrinker, make the next
+	 * one start from the next inode in case it starts before we finish.
+	 */
+	spin_lock(&fs_info->extent_map_shrinker_lock);
+	ctx.last_ino = fs_info->extent_map_shrinker_last_ino;
+	fs_info->extent_map_shrinker_last_ino++;
+	ctx.last_root = fs_info->extent_map_shrinker_last_root;
+	spin_unlock(&fs_info->extent_map_shrinker_lock);
+
+	start_root_id = ctx.last_root;
+	next_root_id = ctx.last_root;
 
 	if (trace_btrfs_extent_map_shrinker_scan_enter_enabled()) {
 		s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
 
-		trace_btrfs_extent_map_shrinker_scan_enter(fs_info, nr_to_scan, nr);
+		trace_btrfs_extent_map_shrinker_scan_enter(fs_info, nr_to_scan,
+							   nr, ctx.last_root,
+							   ctx.last_ino);
 	}
 
-	while (scanned < nr_to_scan) {
+	/*
+	 * We may be called from memory allocation paths, so we don't want to
+	 * take too much time and slowdown tasks, so stop if we need reschedule.
+	 */
+	while (ctx.scanned < ctx.nr_to_scan && !need_resched()) {
 		struct btrfs_root *root;
 		unsigned long count;
 
@@ -1171,8 +1218,8 @@ long btrfs_free_extent_maps(struct btrfs_fs_info *fs_info, long nr_to_scan)
 			spin_unlock(&fs_info->fs_roots_radix_lock);
 			if (start_root_id > 0 && !cycled) {
 				next_root_id = 0;
-				fs_info->extent_map_shrinker_last_root = 0;
-				fs_info->extent_map_shrinker_last_ino = 0;
+				ctx.last_root = 0;
+				ctx.last_ino = 0;
 				cycled = true;
 				continue;
 			}
@@ -1186,15 +1233,33 @@ long btrfs_free_extent_maps(struct btrfs_fs_info *fs_info, long nr_to_scan)
 			continue;
 
 		if (is_fstree(btrfs_root_id(root)))
-			nr_dropped += btrfs_scan_root(root, &scanned, nr_to_scan);
+			nr_dropped += btrfs_scan_root(root, &ctx);
 
 		btrfs_put_root(root);
 	}
 
+	/*
+	 * In case of multiple tasks running this extent map shrinking code this
+	 * isn't perfect but it's simple and silences things like KCSAN. It's
+	 * not possible to know which task made more progress because we can
+	 * cycle back to the first root and first inode if it's not the first
+	 * time the shrinker ran, see the above logic. Also a task that started
+	 * later may finish ealier than another task and made less progress. So
+	 * make this simple and update to the progress of the last task that
+	 * finished, with the occasional possiblity of having two consecutive
+	 * runs of the shrinker process the same inodes.
+	 */
+	spin_lock(&fs_info->extent_map_shrinker_lock);
+	fs_info->extent_map_shrinker_last_ino = ctx.last_ino;
+	fs_info->extent_map_shrinker_last_root = ctx.last_root;
+	spin_unlock(&fs_info->extent_map_shrinker_lock);
+
 	if (trace_btrfs_extent_map_shrinker_scan_exit_enabled()) {
 		s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
 
-		trace_btrfs_extent_map_shrinker_scan_exit(fs_info, nr_dropped, nr);
+		trace_btrfs_extent_map_shrinker_scan_exit(fs_info, nr_dropped,
+							  nr, ctx.last_root,
+							  ctx.last_ino);
 	}
 
 	return nr_dropped;
diff --git a/fs/btrfs/fs.h b/fs/btrfs/fs.h
index 89f0650631cd..833dc3fe0a38 100644
--- a/fs/btrfs/fs.h
+++ b/fs/btrfs/fs.h
@@ -630,6 +630,7 @@ struct btrfs_fs_info {
 	s32 delalloc_batch;
 
 	struct percpu_counter evictable_extent_maps;
+	spinlock_t extent_map_shrinker_lock;
 	u64 extent_map_shrinker_last_root;
 	u64 extent_map_shrinker_last_ino;