selftests/bpf: Reject graph kptr use after RCU unlockgrafted

Add a sleepable verifier test that loads a graph-node local kptr in an
explicit RCU read-side critical section, then passes its node to
bpf_rbtree_remove() after the section ends.

Before the verifier fix, the stale NON_OWN_REF flag makes the node look like
a live borrowed reference and the program is accepted. After the fix, the
pointer is demoted without NON_OWN_REF and the graph kfunc argument is
rejected.

Also exercise a graph kptr loaded while a spin lock provides implicit RCU
protection. The pointer must be invalidated when the lock is released, which
guards the required ordering between non-owning-reference invalidation and
RCU demotion.

Update the existing fault-protected load test state description. The
post-unlock pointer no longer carries NON_OWN_REF, but remains readable
because the load is rewritten to use BPF_PROBE_MEM.

Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20260904084325.52250-7-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

1 files changed, 194 insertions, 0 deletions

diff --git a/fs/omfs/bitmap.c b/fs/omfs/bitmap.c
new file mode 100644
index 000000000..7147ba6a6
--- /dev/null
+++ b/fs/omfs/bitmap.c
@@ -0,0 +1,194 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/kernel.h>
+#include <linux/fs.h>
+#include <linux/buffer_head.h>
+#include <asm/div64.h>
+#include "omfs.h"
+
+unsigned long omfs_count_free(struct super_block *sb)
+{
+	unsigned int i;
+	unsigned long sum = 0;
+	struct omfs_sb_info *sbi = OMFS_SB(sb);
+	int nbits = sb->s_blocksize * 8;
+
+	for (i = 0; i < sbi->s_imap_size; i++)
+		sum += nbits - bitmap_weight(sbi->s_imap[i], nbits);
+
+	return sum;
+}
+
+/*
+ *  Counts the run of zero bits starting at bit up to max.
+ *  It handles the case where a run might spill over a buffer.
+ *  Called with bitmap lock.
+ */
+static int count_run(unsigned long **addr, int nbits,
+		int addrlen, int bit, int max)
+{
+	int count = 0;
+	int x;
+
+	for (; addrlen > 0; addrlen--, addr++) {
+		x = find_next_bit(*addr, nbits, bit);
+		count += x - bit;
+
+		if (x < nbits || count > max)
+			return min(count, max);
+
+		bit = 0;
+	}
+	return min(count, max);
+}
+
+/*
+ * Sets or clears the run of count bits starting with bit.
+ * Called with bitmap lock.
+ */
+static int set_run(struct super_block *sb, int map,
+		int nbits, int bit, int count, int set)
+{
+	int i;
+	int err;
+	struct buffer_head *bh;
+	struct omfs_sb_info *sbi = OMFS_SB(sb);
+
+ 	err = -ENOMEM;
+	bh = sb_bread(sb, clus_to_blk(sbi, sbi->s_bitmap_ino) + map);
+	if (!bh)
+		goto out;
+
+	for (i = 0; i < count; i++, bit++) {
+		if (bit >= nbits) {
+			bit = 0;
+			map++;
+
+			mark_buffer_dirty(bh);
+			brelse(bh);
+			bh = sb_bread(sb,
+				clus_to_blk(sbi, sbi->s_bitmap_ino) + map);
+			if (!bh)
+				goto out;
+		}
+		if (set) {
+			set_bit(bit, sbi->s_imap[map]);
+			set_bit(bit, (unsigned long *)bh->b_data);
+		} else {
+			clear_bit(bit, sbi->s_imap[map]);
+			clear_bit(bit, (unsigned long *)bh->b_data);
+		}
+	}
+	mark_buffer_dirty(bh);
+	brelse(bh);
+	err = 0;
+out:
+	return err;
+}
+
+/*
+ * Tries to allocate exactly one block.  Returns true if successful.
+ */
+int omfs_allocate_block(struct super_block *sb, u64 block)
+{
+	struct buffer_head *bh;
+	struct omfs_sb_info *sbi = OMFS_SB(sb);
+	int bits_per_entry = 8 * sb->s_blocksize;
+	unsigned int map, bit;
+	int ret = 0;
+	u64 tmp;
+
+	tmp = block;
+	bit = do_div(tmp, bits_per_entry);
+	map = tmp;
+
+	mutex_lock(&sbi->s_bitmap_lock);
+	if (map >= sbi->s_imap_size || test_and_set_bit(bit, sbi->s_imap[map]))
+		goto out;
+
+	if (sbi->s_bitmap_ino > 0) {
+		bh = sb_bread(sb, clus_to_blk(sbi, sbi->s_bitmap_ino) + map);
+		if (!bh)
+			goto out;
+
+		set_bit(bit, (unsigned long *)bh->b_data);
+		mark_buffer_dirty(bh);
+		brelse(bh);
+	}
+	ret = 1;
+out:
+	mutex_unlock(&sbi->s_bitmap_lock);
+	return ret;
+}
+
+
+/*
+ *  Tries to allocate a set of blocks.	The request size depends on the
+ *  type: for inodes, we must allocate sbi->s_mirrors blocks, and for file
+ *  blocks, we try to allocate sbi->s_clustersize, but can always get away
+ *  with just one block.
+ */
+int omfs_allocate_range(struct super_block *sb,
+			int min_request,
+			int max_request,
+			u64 *return_block,
+			int *return_size)
+{
+	struct omfs_sb_info *sbi = OMFS_SB(sb);
+	int bits_per_entry = 8 * sb->s_blocksize;
+	int ret = 0;
+	int i, run, bit;
+
+	mutex_lock(&sbi->s_bitmap_lock);
+	for (i = 0; i < sbi->s_imap_size; i++) {
+		bit = 0;
+		while (bit < bits_per_entry) {
+			bit = find_next_zero_bit(sbi->s_imap[i], bits_per_entry,
+				bit);
+
+			if (bit == bits_per_entry)
+				break;
+
+			run = count_run(&sbi->s_imap[i], bits_per_entry,
+				sbi->s_imap_size-i, bit, max_request);
+
+			if (run >= min_request)
+				goto found;
+			bit += run;
+		}
+	}
+	ret = -ENOSPC;
+	goto out;
+
+found:
+	*return_block = (u64) i * bits_per_entry + bit;
+	*return_size = run;
+	ret = set_run(sb, i, bits_per_entry, bit, run, 1);
+
+out:
+	mutex_unlock(&sbi->s_bitmap_lock);
+	return ret;
+}
+
+/*
+ * Clears count bits starting at a given block.
+ */
+int omfs_clear_range(struct super_block *sb, u64 block, int count)
+{
+	struct omfs_sb_info *sbi = OMFS_SB(sb);
+	int bits_per_entry = 8 * sb->s_blocksize;
+	u64 tmp;
+	unsigned int map, bit;
+	int ret;
+
+	tmp = block;
+	bit = do_div(tmp, bits_per_entry);
+	map = tmp;
+
+	if (map >= sbi->s_imap_size)
+		return 0;
+
+	mutex_lock(&sbi->s_bitmap_lock);
+	ret = set_run(sb, map, bits_per_entry, bit, count, 0);
+	mutex_unlock(&sbi->s_bitmap_lock);
+	return ret;
+}