linux-kernel - [PATCH] RAS/AMD/ATL: Replace bitwise_xor

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20251124142517.1708451-1-nik.borisov@suse.com>
Date: Mon, 24 Nov 2025 16:25:17 +0200
From: Nikolay Borisov <nik.borisov@...e.com>
To: linux-edac@...r.kernel.org
Cc: linux-kernel@...r.kernel.org,
	bp@...en8.de,
	Yazen.Ghannam@....com,
	Nikolay Borisov <nik.borisov@...e.com>
Subject: [PATCH] RAS/AMD/ATL: Replace bitwise_xor_bits() with hweight16()

Doing hweight16 and checking whether the lsb is set is functionally
equivalent to what bitwise_xor_bits() does. In addition it results in
better generated code as before gcc would inline the function 4 times.
With hweight, the resulting code boils down to 2 instructions -  popcnt
and andl as all cpus we care about has popcnt. No functional changes.

An alternative would have been to use the __builtin_parity() function provided
by both Clang/GCC, however under some circumstances the compiler can choose not
to inline it but generate a library call which is unsupported in the kernel.

Signed-off-by: Nikolay Borisov <nik.borisov@...e.com>
---
 drivers/ras/amd/atl/umc.c | 21 +++++----------------
 1 file changed, 5 insertions(+), 16 deletions(-)

diff --git a/drivers/ras/amd/atl/umc.c b/drivers/ras/amd/atl/umc.c
index 6e072b7667e9..8d1363237eee 100644
--- a/drivers/ras/amd/atl/umc.c
+++ b/drivers/ras/amd/atl/umc.c
@@ -49,17 +49,6 @@ static u8 get_coh_st_inst_id_mi300(struct atl_err *err)
 	return i;
 }

-/* XOR the bits in @val. */
-static u16 bitwise_xor_bits(u16 val)
-{
-	u16 tmp = 0;
-	u8 i;
-
-	for (i = 0; i < 16; i++)
-		tmp ^= (val >> i) & 0x1;
-
-	return tmp;
-}

 struct xor_bits {
 	bool	xor_enable;
@@ -250,17 +239,17 @@ static unsigned long convert_dram_to_norm_addr_mi300(unsigned long addr)
 		if (!addr_hash.bank[i].xor_enable)
 			continue;

-		temp  = bitwise_xor_bits(col & addr_hash.bank[i].col_xor);
-		temp ^= bitwise_xor_bits(row & addr_hash.bank[i].row_xor);
+		temp  = hweight16(col & addr_hash.bank[i].col_xor) & 1;
+		temp ^= hweight16(row & addr_hash.bank[i].row_xor) & 1;
 		bank ^= temp << i;
 	}

 	/* Calculate hash for PC bit. */
 	if (addr_hash.pc.xor_enable) {
-		temp  = bitwise_xor_bits(col  & addr_hash.pc.col_xor);
-		temp ^= bitwise_xor_bits(row  & addr_hash.pc.row_xor);
+		temp  = hweight16(col & addr_hash.pc.col_xor) & 1;
+		temp ^= hweight16(row & addr_hash.pc.row_xor) & 1;
 		/* Bits SID[1:0] act as Bank[5:4] for PC hash, so apply them here. */
-		temp ^= bitwise_xor_bits((bank | sid << NUM_BANK_BITS) & addr_hash.bank_xor);
+		temp ^= hweight16((bank | sid << NUM_BANK_BITS) & addr_hash.bank_xor) & 1;
 		pc   ^= temp;
 	}

--
2.52.0