x86/amd_node: Fix PCI device reference counting in amd_smn_init()grafted

The local "root" pointer is a temporary variable used during the device
search. Therefore, refcount related to the search iterators should be cleaned
up after the search is complete.

Use the __free() cleanup macro to ensure the refcount is decremented when the
temporary pointer goes out of scope.

Additionally, increment the refcount when caching a root pointer. This ensures
the in-use refcount is separate from the temporary search refcounting.

Finally, drop the redundant "root = NULL" before the second search loop. The
pci_get_class() iterator always decrements the refcount of its "from"
argument, so the first loop can only fall through with "root" already NULL.

Fixes: 0a4b61d9c2e4 ("x86/amd_node: Fix AMD root device caching")
Closes: https://sashiko.dev/#/patchset/20260806160159.230453-1-jason.andryuk%40amd.com
Reported-by: Sashiko <sashiko-bot@kernel.org>
Assisted-by: LLM
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Cc: <stable@kernel.org>
Link: https://patch.msgid.link/20260903154325.74343-1-yazen.ghannam@amd.com

1 files changed, 67 insertions, 0 deletions

diff --git a/arch/arc/lib/strcpy-700.S b/arch/arc/lib/strcpy-700.S
new file mode 100644
index 000000000..6e2294d13
--- /dev/null
+++ b/arch/arc/lib/strcpy-700.S
@@ -0,0 +1,67 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2004, 2007-2010, 2011-2012 Synopsys, Inc. (www.synopsys.com)
+ */
+
+/* If dst and src are 4 byte aligned, copy 8 bytes at a time.
+   If the src is 4, but not 8 byte aligned, we first read 4 bytes to get
+   it 8 byte aligned.  Thus, we can do a little read-ahead, without
+   dereferencing a cache line that we should not touch.
+   Note that short and long instructions have been scheduled to avoid
+   branch stalls.
+   The beq_s to r3z could be made unaligned & long to avoid a stall
+   there, but the it is not likely to be taken often, and it
+   would also be likey to cost an unaligned mispredict at the next call.  */
+
+#include <linux/linkage.h>
+
+ENTRY_CFI(strcpy)
+	or	r2,r0,r1
+	bmsk_s	r2,r2,1
+	brne.d	r2,0,charloop
+	mov_s	r10,r0
+	ld_s	r3,[r1,0]
+	mov	r8,0x01010101
+	bbit0.d	r1,2,loop_start
+	ror	r12,r8
+	sub	r2,r3,r8
+	bic_s	r2,r2,r3
+	tst_s	r2,r12
+	bne	r3z
+	mov_s	r4,r3
+	.balign 4
+loop:
+	ld.a	r3,[r1,4]
+	st.ab	r4,[r10,4]
+loop_start:
+	ld.a	r4,[r1,4]
+	sub	r2,r3,r8
+	bic_s	r2,r2,r3
+	tst_s	r2,r12
+	bne_s	r3z
+	st.ab	r3,[r10,4]
+	sub	r2,r4,r8
+	bic	r2,r2,r4
+	tst	r2,r12
+	beq	loop
+	mov_s	r3,r4
+#ifdef __LITTLE_ENDIAN__
+r3z:	bmsk.f	r1,r3,7
+	lsr_s	r3,r3,8
+#else
+r3z:	lsr.f	r1,r3,24
+	asl_s	r3,r3,8
+#endif
+	bne.d	r3z
+	stb.ab	r1,[r10,1]
+	j_s	[blink]
+
+	.balign	4
+charloop:
+	ldb.ab	r3,[r1,1]
+
+
+	brne.d	r3,0,charloop
+	stb.ab	r3,[r10,1]
+	j	[blink]
+END_CFI(strcpy)