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, 75 insertions, 0 deletions

diff --git a/arch/xtensa/lib/divsi3.S b/arch/xtensa/lib/divsi3.S
new file mode 100644
index 000000000..edb3c4ad9
--- /dev/null
+++ b/arch/xtensa/lib/divsi3.S
@@ -0,0 +1,75 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later WITH GCC-exception-2.0 */
+#include <linux/linkage.h>
+#include <asm/asmmacro.h>
+#include <asm/core.h>
+
+ENTRY(__divsi3)
+
+	abi_entry_default
+#if XCHAL_HAVE_DIV32
+	quos	a2, a2, a3
+#else
+	xor	a7, a2, a3	/* sign = dividend ^ divisor */
+	do_abs	a6, a2, a4	/* udividend = abs (dividend) */
+	do_abs	a3, a3, a4	/* udivisor = abs (divisor) */
+	bltui	a3, 2, .Lle_one	/* check if udivisor <= 1 */
+	do_nsau	a5, a6, a2, a8	/* udividend_shift = nsau (udividend) */
+	do_nsau	a4, a3, a2, a8	/* udivisor_shift = nsau (udivisor) */
+	bgeu	a5, a4, .Lspecial
+
+	sub	a4, a4, a5	/* count = udivisor_shift - udividend_shift */
+	ssl	a4
+	sll	a3, a3		/* udivisor <<= count */
+	movi	a2, 0		/* quotient = 0 */
+
+	/* test-subtract-and-shift loop; one quotient bit on each iteration */
+#if XCHAL_HAVE_LOOPS
+	loopnez	a4, .Lloopend
+#endif /* XCHAL_HAVE_LOOPS */
+.Lloop:
+	bltu	a6, a3, .Lzerobit
+	sub	a6, a6, a3
+	addi	a2, a2, 1
+.Lzerobit:
+	slli	a2, a2, 1
+	srli	a3, a3, 1
+#if !XCHAL_HAVE_LOOPS
+	addi	a4, a4, -1
+	bnez	a4, .Lloop
+#endif /* !XCHAL_HAVE_LOOPS */
+.Lloopend:
+
+	bltu	a6, a3, .Lreturn
+	addi	a2, a2, 1	/* increment if udividend >= udivisor */
+.Lreturn:
+	neg	a5, a2
+	movltz	a2, a5, a7	/* return (sign < 0) ? -quotient : quotient */
+	abi_ret_default
+
+.Lle_one:
+	beqz	a3, .Lerror
+	neg	a2, a6		/* if udivisor == 1, then return... */
+	movgez	a2, a6, a7	/* (sign < 0) ? -udividend : udividend */
+	abi_ret_default
+
+.Lspecial:
+	bltu	a6, a3, .Lreturn0 /* if dividend < divisor, return 0 */
+	movi	a2, 1
+	movi	a4, -1
+	movltz	a2, a4, a7	/* else return (sign < 0) ? -1 : 1 */
+	abi_ret_default
+
+.Lerror:
+	/* Divide by zero: Use an illegal instruction to force an exception.
+	   The subsequent "DIV0" string can be recognized by the exception
+	   handler to identify the real cause of the exception.  */
+	ill
+	.ascii	"DIV0"
+
+.Lreturn0:
+	movi	a2, 0
+#endif /* XCHAL_HAVE_DIV32 */
+	abi_ret_default
+
+ENDPROC(__divsi3)
+EXPORT_SYMBOL(__divsi3)