Skip to contentSkip to conversation

X86: MM: Add PAT Type write-through in combination with mtrr

PATCH In review Andreas Werner 2 participants 4 messages Nov 3, 2013 – Nov 19, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 X86: MM: Add PAT Type write-through in combination with mtrr

Revision 2:
added comment in code.

This patch adds the Write-through memory type in combination with mtrr. If you call ioremap_cache to request cachable memory (write-back) the function tries to set the PAT to write-back only if the mtrr setting of the requested region is also marked as Write-Back.

If the mttr regions are marked e.g. as Write-through or with other types, the function will always return UC- memory.

If you check the Intel document " IA-32 SDM vol 3a table Effective Memory Type", there
are many other combinations possible.

This patch will only add the following combination:
PAT=Write-Back + MTRR=Write-Through.

Since marking IO Memory as cachable is not valid, WT is the best way for caching/bursting on MMIO Devices.

Tested on - Intel (R) Atom E680 (Tunnel Creek)
- Intel (R) Core(TM)2 Duo

---
 arch/x86/mm/pat.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)
arch/x86/mm/pat.c +15 -3
index 6574388..cf05ee6 100644
@@ -149,10 +149,22 @@ static unsigned long pat_x_mtrr_type(u64 start, u64 end, unsigned long req_type)
 		u8 mtrr_type;
 
 		mtrr_type = mtrr_type_lookup(start, end);
-		if (mtrr_type != MTRR_TYPE_WRBACK)
-			return _PAGE_CACHE_UC_MINUS;
 
-		return _PAGE_CACHE_WB;
+		switch (mtrr_type) {
+		case MTRR_TYPE_WRBACK:
+
+		/*
+		 * Return also WB (PAT) if MTTR is set to WT.
+		 * Since marking IO Memory as cachable is not valid,
+		 * WT is the best way for caching/bursting reads
+		 * on MMIO.
+		 */
+		case MTRR_TYPE_WRTHROUGH:
+		  return _PAGE_CACHE_WB;
+
+		default:
+		  return _PAGE_CACHE_UC_MINUS;
+		}
 	}
 
 	return req_type;
Signature · 1 line
1.8.4
#1 lore raw
Quoted 4 lines · Andreas Werner

Since marking IO Memory as cachable is not valid, WT is the best way for caching/bursting on MMIO Devices.

Uh... WT *is* cachable...

-hpa

#2 lore raw
Quoted 11 lines · H. Peter Anvin
Quoted 4 lines · Andreas Werner

Since marking IO Memory as cachable is not valid, WT is the best way for caching/bursting on MMIO Devices.

Uh... WT *is* cachable...

-hpa

Sorry my mistake, i mean "marking IO Memory as WB is not valid"

regards
Andy

#3 lore raw
In reply to Andreas Werner
Quoted 18 lines · Andreas Werner

Revision 2:
added comment in code.

This patch adds the Write-through memory type in combination with mtrr. If you call ioremap_cache to request cachable memory (write-back) the function tries to set the PAT to write-back only if the mtrr setting of the requested region is also marked as Write-Back.

If the mttr regions are marked e.g. as Write-through or with other types, the function will always return UC- memory.

If you check the Intel document " IA-32 SDM vol 3a table Effective Memory Type", there
are many other combinations possible.

This patch will only add the following combination:
PAT=Write-Back + MTRR=Write-Through.

Quoted 7 lines

Since marking IO Memory as cachable is not valid, WT is the best way for caching/bursting on MMIO Devices.

Tested on - Intel (R) Atom E680 (Tunnel Creek)
- Intel (R) Core(TM)2 Duo

I don't quite know where this ended up, but I am *really* not happy about going back to using MTRRs to mark I/O devices with the chronic problems of MTRR exhaustion that entails. As such I do insist that PAT is properly updated to support WT if we're going to do this.

-hpa