public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH V2] x86/gart: Check for GART support before accessing GART registers
@ 2015-04-07 21:46 Aravind Gopalakrishnan
  2015-04-08 15:25 ` Borislav Petkov
  2015-05-06 10:13 ` [tip:x86/cpu] " tip-bot for Aravind Gopalakrishnan
  0 siblings, 2 replies; 3+ messages in thread
From: Aravind Gopalakrishnan @ 2015-04-07 21:46 UTC (permalink / raw)
  To: mingo, bp, linux-kernel
  Cc: Suravee.Suthikulpanit, Joerg Rodel, Thomas Gleixner, Ingo Molnar,
	H. Peter Anvin, Bjorn Helgaas, x86-ml

GART registers are not present in newer processors (Fam15h, Model 10h
and later). So, avoid accesses to GART registers in PCI config
space by returning early in early_gart_iommu_check() and
gart_iommu_hole_init() if GART is not available.

Current code doesn't break on existing processors but there are some
side effects:

We get bogus AGP aperture messages which are simply noise on
GART-less processors:

  AGP: Node 0: aperture [bus addr 0x00000000-0x01ffffff] (32MB)
  AGP: Your BIOS doesn't leave aperture memory hole
  AGP: Please enable the IOMMU option in the BIOS setup
  AGP: This costs you 64MB of RAM
  AGP: Mapping aperture over RAM [mem 0xd4000000-0xd7ffffff]

We can avoid calling allocate_aperture() and would not have to
wastefully reserve 64MB of RAM with memblock_reserve(). Also, we can
avoid having to loop through all PCI buses and devices twice, searching
for a non-existent AGP bridge if we bail out early.

Refactoring the family check used in amd_nb.c into an inline function
so we can use it here as well as in amd_nb.c

Fix some typos while at it.
Tested the patch on Fam10h and Fam15h Model 00h-fh and this code
runs fine. On Fam15h Model 60h-6fh and on Fam16h, we bail early
as they don't have GART.

Signed-off-by: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>
Reviewed-by: Suravee Suthikulpanit <Suravee.Suthikulpanit@amd.com>
Cc: Joerg Rodel <joro@8bytes.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: x86-ml <x86@kernel.org>
---
Changes in V2:
 - move amd_gart_present() before any other checks as it's least intrusive (per Boris)
 - fix typos in code and commit message (per Ingo)
 - minor change to subject line (per Boris' change)
 - Link to V1:http://marc.info/?l=linux-kernel&m=142792069809730&w=2

 arch/x86/include/asm/amd_nb.h | 11 +++++++++++
 arch/x86/kernel/amd_nb.c      |  4 +---
 arch/x86/kernel/aperture_64.c |  8 +++++++-
 3 files changed, 19 insertions(+), 4 deletions(-)

diff --git a/arch/x86/include/asm/amd_nb.h b/arch/x86/include/asm/amd_nb.h
index aaac3b2..1a5da2e 100644
--- a/arch/x86/include/asm/amd_nb.h
+++ b/arch/x86/include/asm/amd_nb.h
@@ -98,11 +98,22 @@ static inline u16 amd_get_node_id(struct pci_dev *pdev)
 	return 0;
 }
 
+static inline bool amd_gart_present(void)
+{
+	/* GART present only on Fam15h, upto model 0fh */
+	if (boot_cpu_data.x86 == 0xf || boot_cpu_data.x86 == 0x10 ||
+	    (boot_cpu_data.x86 == 0x15 && boot_cpu_data.x86_model < 0x10))
+		return true;
+
+	return false;
+}
+
 #else
 
 #define amd_nb_num(x)		0
 #define amd_nb_has_feature(x)	false
 #define node_to_amd_nb(x)	NULL
+#define amd_gart_present(x)	false
 
 #endif
 
diff --git a/arch/x86/kernel/amd_nb.c b/arch/x86/kernel/amd_nb.c
index 5caed1d..29fa475 100644
--- a/arch/x86/kernel/amd_nb.c
+++ b/arch/x86/kernel/amd_nb.c
@@ -89,9 +89,7 @@ int amd_cache_northbridges(void)
 			next_northbridge(link, amd_nb_link_ids);
 	}
 
-	/* GART present only on Fam15h upto model 0fh */
-	if (boot_cpu_data.x86 == 0xf || boot_cpu_data.x86 == 0x10 ||
-	    (boot_cpu_data.x86 == 0x15 && boot_cpu_data.x86_model < 0x10))
+	if (amd_gart_present())
 		amd_northbridges.flags |= AMD_NB_GART;
 
 	/*
diff --git a/arch/x86/kernel/aperture_64.c b/arch/x86/kernel/aperture_64.c
index 76164e1..5c3e9b56 100644
--- a/arch/x86/kernel/aperture_64.c
+++ b/arch/x86/kernel/aperture_64.c
@@ -262,6 +262,9 @@ void __init early_gart_iommu_check(void)
 	u64 aper_base = 0, last_aper_base = 0;
 	int aper_enabled = 0, last_aper_enabled = 0, last_valid = 0;
 
+	if (!amd_gart_present())
+		return;
+
 	if (!early_pci_allowed())
 		return;
 
@@ -355,6 +358,9 @@ int __init gart_iommu_hole_init(void)
 	int fix, slot, valid_agp = 0;
 	int i, node;
 
+	if (!amd_gart_present())
+		return -ENODEV;
+
 	if (gart_iommu_aperture_disabled || !fix_aperture ||
 	    !early_pci_allowed())
 		return -ENODEV;
@@ -452,7 +458,7 @@ out:
 		   force_iommu ||
 		   valid_agp ||
 		   fallback_aper_force) {
-		pr_info("Your BIOS doesn't leave a aperture memory hole\n");
+		pr_info("Your BIOS doesn't leave aperture memory hole\n");
 		pr_info("Please enable the IOMMU option in the BIOS setup\n");
 		pr_info("This costs you %dMB of RAM\n",
 			32 << fallback_aper_order);
-- 
1.9.1


^ permalink raw reply related	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2015-05-06 10:14 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-04-07 21:46 [PATCH V2] x86/gart: Check for GART support before accessing GART registers Aravind Gopalakrishnan
2015-04-08 15:25 ` Borislav Petkov
2015-05-06 10:13 ` [tip:x86/cpu] " tip-bot for Aravind Gopalakrishnan

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox