[RFC][PATCH 19/34] x86/cpu: Introduce cache alignment multiplier

[Dave Hansen <dave.hansen@linux.intel.com>]

From: Dave Hansen <dave.hansen@linux.intel.com>

The kernel currently tracks two different but similar-sounding values:

	x86_clflush_size
and
	x86_cache_alignment

x86_clflush_size is literally the size of the cacheline which is also
the architectural granularity at which CLFLUSH operates.

But some weirdo CPUs like the Pentium 4 do some operations across
two cachelines.  Their CLFLUSH still works at 'x86_clflush_size'
but they need some operations aligned to that double cacheline to
work well, thus 'x86_cache_alignment'.

Introduce and use a 'bsp_addr_config' field for these systems.

Note that the Intel Family 15 code actually did this in two different
sites, one for 32-bit and one for 64-bit.  Unify them.

Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
---

 b/arch/x86/include/asm/processor.h |    7 +++++++
 b/arch/x86/kernel/cpu/centaur.c    |    2 +-
 b/arch/x86/kernel/cpu/common.c     |    3 +++
 b/arch/x86/kernel/cpu/intel.c      |   10 ++++------
 4 files changed, 15 insertions(+), 7 deletions(-)

diff -puN arch/x86/include/asm/processor.h~x86_cache_alignment_mult arch/x86/include/asm/processor.h
--- a/arch/x86/include/asm/processor.h~x86_cache_alignment_mult	2024-02-22 10:08:57.732817356 -0800
+++ b/arch/x86/include/asm/processor.h	2024-02-22 10:08:57.740817669 -0800
@@ -172,6 +172,13 @@ struct x86_addr_config {
 	 * platforms that use memory encryption.
 	 */
 	u8 phys_addr_reduction_bits;
+
+	/*
+	 * "x86_clflush_size" is the size of an actual cacheline.
+	 * Allow systems to specify a multiplier where alignment
+	 * will take place at a more coarse granularity.
+	 */
+	u8 cache_align_mult;
 };
 
 /*
diff -puN arch/x86/kernel/cpu/centaur.c~x86_cache_alignment_mult arch/x86/kernel/cpu/centaur.c
--- a/arch/x86/kernel/cpu/centaur.c~x86_cache_alignment_mult	2024-02-22 10:08:57.732817356 -0800
+++ b/arch/x86/kernel/cpu/centaur.c	2024-02-22 10:08:57.740817669 -0800
@@ -216,7 +216,7 @@ static void init_centaur(struct cpuinfo_
 static void bsp_init_centaur(struct cpuinfo_x86 *c)
 {
 	if (c->x86 == 0x6 && c->x86_model >= 0xf)
-		c->x86_cache_alignment = x86_clflush_size() * 2;
+		bsp_addr_config.cache_align_mult = 2;
 }
 
 #ifdef CONFIG_X86_32
diff -puN arch/x86/kernel/cpu/common.c~x86_cache_alignment_mult arch/x86/kernel/cpu/common.c
--- a/arch/x86/kernel/cpu/common.c~x86_cache_alignment_mult	2024-02-22 10:08:57.736817513 -0800
+++ b/arch/x86/kernel/cpu/common.c	2024-02-22 10:08:57.740817669 -0800
@@ -1124,7 +1124,10 @@ void get_cpu_address_sizes(struct cpuinf
 		}
 	}
 	c->x86_cache_bits = x86_config.phys_bits;
+
 	c->x86_cache_alignment = x86_clflush_size();
+	if (bsp_addr_config.cache_align_mult)
+		c->x86_cache_alignment *= bsp_addr_config.cache_align_mult;
 
 	/* Do this last to avoid affecting ->x86_cache_bits. */
 	x86_config.phys_bits -= bsp_addr_config.phys_addr_reduction_bits;
diff -puN arch/x86/kernel/cpu/intel.c~x86_cache_alignment_mult arch/x86/kernel/cpu/intel.c
--- a/arch/x86/kernel/cpu/intel.c~x86_cache_alignment_mult	2024-02-22 10:08:57.736817513 -0800
+++ b/arch/x86/kernel/cpu/intel.c	2024-02-22 10:08:57.740817669 -0800
@@ -236,10 +236,6 @@ static void early_init_intel(struct cpui
 
 #ifdef CONFIG_X86_64
 	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#else
-	/* Netburst reports 64 bytes clflush size, but does IO in 128 bytes */
-	if (c->x86 == 15 && c->x86_cache_alignment == 64)
-		c->x86_cache_alignment = 128;
 #endif
 
 	/*
@@ -418,6 +414,10 @@ static void bsp_init_intel(struct cpuinf
 		WARN_ON_ONCE(keyid_bits);
 		bsp_addr_config.phys_addr_reduction_bits = 4;
 	}
+
+	/* Netburst reports 64 bytes clflush size, but does IO in 128 bytes */
+	if (c->x86 == 15)
+		bsp_addr_config.cache_align_mult = 2;
 }
 
 #ifdef CONFIG_X86_32
@@ -659,8 +659,6 @@ static void init_intel(struct cpuinfo_x8
 		set_cpu_bug(c, X86_BUG_MONITOR);
 
 #ifdef CONFIG_X86_64
-	if (c->x86 == 15)
-		c->x86_cache_alignment = x86_clflush_size() * 2;
 	if (c->x86 == 6)
 		set_cpu_cap(c, X86_FEATURE_REP_GOOD);
 #else
_