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, 96 insertions, 0 deletions

diff --git a/arch/m68k/mm/hwtest.c b/arch/m68k/mm/hwtest.c
new file mode 100644
index 000000000..8ee7a3368
--- /dev/null
+++ b/arch/m68k/mm/hwtest.c
@@ -0,0 +1,96 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Tests for presence or absence of hardware registers.
+ * This code was originally in atari/config.c, but I noticed
+ * that it was also in drivers/nubus/nubus.c and I wanted to
+ * use it in hp300/config.c, so it seemed sensible to pull it
+ * out into its own file.
+ *
+ * The test is for use when trying to read a hardware register
+ * that isn't present would cause a bus error. We set up a
+ * temporary handler so that this doesn't kill the kernel.
+ *
+ * There is a test-by-reading and a test-by-writing; I present
+ * them here complete with the comments from the original atari
+ * config.c...
+ *                -- PMM <pmaydell@chiark.greenend.org.uk>, 05/1998
+ */
+
+/* This function tests for the presence of an address, specially a
+ * hardware register address. It is called very early in the kernel
+ * initialization process, when the VBR register isn't set up yet. On
+ * an Atari, it still points to address 0, which is unmapped. So a bus
+ * error would cause another bus error while fetching the exception
+ * vector, and the CPU would do nothing at all. So we needed to set up
+ * a temporary VBR and a vector table for the duration of the test.
+ */
+
+#include <linux/module.h>
+
+#include <asm/hwtest.h>
+
+int hwreg_present(volatile void *regp)
+{
+	int ret = 0;
+	unsigned long flags;
+	long save_sp, save_vbr;
+	long tmp_vectors[3];
+
+	local_irq_save(flags);
+	__asm__ __volatile__ (
+		"movec %/vbr,%2\n\t"
+		"movel #Lberr1,%4@(8)\n\t"
+		"movec %4,%/vbr\n\t"
+		"movel %/sp,%1\n\t"
+		"moveq #0,%0\n\t"
+		"tstb %3@\n\t"
+		"nop\n\t"
+		"moveq #1,%0\n"
+	"Lberr1:\n\t"
+		"movel %1,%/sp\n\t"
+		"movec %2,%/vbr"
+		: "=&d" (ret), "=&r" (save_sp), "=&r" (save_vbr)
+		: "a" (regp), "a" (tmp_vectors)
+	);
+	local_irq_restore(flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(hwreg_present);
+
+/* Basically the same, but writes a value into a word register, protected
+ * by a bus error handler. Returns 1 if successful, 0 otherwise.
+ */
+
+int hwreg_write(volatile void *regp, unsigned short val)
+{
+	int ret;
+	unsigned long flags;
+	long save_sp, save_vbr;
+	long tmp_vectors[3];
+
+	local_irq_save(flags);
+	__asm__ __volatile__ (
+		"movec %/vbr,%2\n\t"
+		"movel #Lberr2,%4@(8)\n\t"
+		"movec %4,%/vbr\n\t"
+		"movel %/sp,%1\n\t"
+		"moveq #0,%0\n\t"
+		"movew %5,%3@\n\t"
+		"nop\n\t"
+		/*
+		 * If this nop isn't present, 'ret' may already be loaded
+		 * with 1 at the time the bus error happens!
+		 */
+		"moveq #1,%0\n"
+	"Lberr2:\n\t"
+		"movel %1,%/sp\n\t"
+		"movec %2,%/vbr"
+		: "=&d" (ret), "=&r" (save_sp), "=&r" (save_vbr)
+		: "a" (regp), "a" (tmp_vectors), "g" (val)
+	);
+	local_irq_restore(flags);
+
+	return ret;
+}
+EXPORT_SYMBOL(hwreg_write);
+