x86, aperture: Check for GART before accessing GART registers
PATCH 1/1 x86, aperture: Check for GART before accessing GART registers
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.
Refactoring the family check used in amd_nb into an inline function so we can use it here as well as in amd_nb.c
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.
--- arch/x86/include/asm/amd_nb.h | 11 +++++++++++ arch/x86/kernel/amd_nb.c | 4 +--- arch/x86/kernel/aperture_64.c | 4 ++-- 3 files changed, 14 insertions(+), 5 deletions(-)
arch/x86/include/asm/amd_nb.h +11 -0
@@ -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
arch/x86/kernel/amd_nb.c +1 -3
@@ -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 ;/*
arch/x86/kernel/aperture_64.c +2 -2
@@ -262,7 +262,7 @@ 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 (!early_pci_allowed ()) +if (!early_pci_allowed () || !amd_gart_present ())return ;/* This is mostly duplicate of iommu_hole_init */
@@ -356,7 +356,7 @@ int __init gart_iommu_hole_init(void)int i, node;if (gart_iommu_aperture_disabled || !fix_aperture || - !early_pci_allowed ()) + !early_pci_allowed () || !amd_gart_present ())return -ENODEV ;pr_info ("Checking aperture...\n" );
Signature · 1 line
1.9.1
Borislav Petkov
AppliedQuoted 12 lines · Aravind Gopalakrishnan
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()andgart_iommu_hole_init()if GART is not available.Refactoring the family check used in
amd_nbinto an inline function so we can use it here as well as inamd_nb.cTested 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.
Quoted 7 lines
Signed-off-by Aravind Gopalakrishnan <[email protected]>Reviewed-by Suravee Suthikulpanit <[email protected]>--- arch/x86/include/asm/amd_nb.h | 11 +++++++++++ arch/x86/kernel/amd_nb.c | 4 +--- arch/x86/kernel/aperture_64.c | 4 ++-- 3 files changed, 14 insertions(+), 5 deletions(-)
Applied, thanks.
Signature · 5 lines
Regards/Gruss,
Boris.
ECO tip #101: Trim your mails when you reply.
--Quoted 4 lines · Aravind Gopalakrishnan
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()andgart_iommu_hole_init()if GART is not available.
In what fashion did this problem manifest itself on real systems?
Thanks,
Ingo
Quoted 9 lines · Ingo Molnar
Quoted 4 lines · Aravind Gopalakrishnan
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()andgart_iommu_hole_init()if GART is not available.In what fashion did this problem manifest itself on real systems?
This code doesn't break on existing processors.
There are some other side effects though..
We get "AGP:" messages on kernel logs like this-
[ 0.000000] AGP: Node 0: aperture [bus addr 0x00000000-0x01ffffff] (32MB) [ 0.000000] AGP: Your BIOS doesn't leave a aperture memory hole [ 0.000000] AGP: Please enable the IOMMU option in the BIOS setup [ 0.000000] AGP: This costs you 64MB of RAM [ 0.000000] AGP: Mapping aperture over RAM [mem 0xd4000000-0xd7ffffff] (65536KB)
These are just noise on processors which have no GART.
We can avoid calling allocate_aperture() and would not have to memblock_reserve() 64MB of RAM.
Also, we can avoid having to loop through all PCI buses, devices (twice) searching for AGP bridge if we bail out early.
For future processors though, we can't say how these registers will be implemented.
If we have to provide PCI IDs for them in amd_nb_misc_ids[] and this code ends up reading/writing something else then that'd be another problem.
Thanks,
-Aravind.
Quoted 15 lines · Aravind Gopalakrishnan
Quoted 9 lines · Ingo Molnar
Quoted 4 lines · Aravind Gopalakrishnan
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()andgart_iommu_hole_init()if GART is not available.In what fashion did this problem manifest itself on real systems?
This code doesn't break on existing processors.
There are some other side effects though..We get "AGP:" messages on kernel logs like this-
Quoted 8 lines
[ 0.000000] AGP: Node 0: aperture [bus addr 0x00000000-0x01ffffff] (32MB) [ 0.000000] AGP: Your BIOS doesn't leave a aperture memory hole [ 0.000000] AGP: Please enable the IOMMU option in the BIOS setup [ 0.000000] AGP: This costs you 64MB of RAM [ 0.000000] AGP: Mapping aperture over RAM [mem 0xd4000000-0xd7ffffff](65536KB)
These are just noise on processors which have no GART.
agreed.
Quoted 4 lines
We can avoid calling
allocate_aperture()and would not have tomemblock_reserve()64MB of RAM.
Also, we can avoid having to loop through all PCI buses, devices (twice) searching for AGP bridge if we bail out early.
Makes sense. Mind adding this info to the changelog and resend?
Thanks,
Ingo
Quoted 10 lines · Ingo Molnar
We get "AGP:" messages on kernel logs like this-
[ 0.000000] AGP: Node 0: aperture [bus addr 0x00000000-0x01ffffff] (32MB) [ 0.000000] AGP: Your BIOS doesn't leave a aperture memory hole [ 0.000000] AGP: Please enable the IOMMU option in the BIOS setup [ 0.000000] AGP: This costs you 64MB of RAM [ 0.000000] AGP: Mapping aperture over RAM [mem 0xd4000000-0xd7ffffff](65536KB)
Quoted 10 lines
These are just noise on processors which have no GART. agreed.
Quoted 4 lines
We can avoid calling
allocate_aperture()and would not have tomemblock_reserve()64MB of RAM.
Also, we can avoid having to loop through all PCI buses, devices (twice) searching for AGP bridge if we bail out early.Makes sense. Mind adding this info to the changelog and resend?
Sure, will do that and resend.
Thanks,
-Aravind.
Quoted 1 line · Aravind Gopalakrishnan
Sure, will do that and resend.
No need - I can amend the local copy I have here.
Signature · 5 lines
Regards/Gruss,
Boris.
ECO tip #101: Trim your mails when you reply.
--Quoted 4 lines · Borislav Petkov
Quoted 1 line · Aravind Gopalakrishnan
Sure, will do that and resend.
No need - I can amend the local copy I have here.
Okay.
Thanks!
-Aravind.
Borislav Petkov
Reviewed-by Suravee SuthikulpanitQuoted 1 line · Aravind Gopalakrishnan
Quoted 1 line
No need - I can amend the local copy I have here.
Here's what I did:
---
From: Aravind Gopalakrishnan <[email protected]>
Date: Wed, 1 Apr 2015 09:32:08 -0500
Subject: [PATCH] x86/gart: Check for GART support before accessing GART
registers
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
GART registers are not present in newer AMD 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 a 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.
Refactor the family check used in amd_nb into an inline function so we can use it here as well as in amd_nb.c.
Tested the patch on Fam10h and Fam15h Model 00h-fh and this code runs fine. On Fam15h, Models 60h-6fh and on Fam16h, we bail out early as they don't have a GART.
Signed-off-by:
--- arch/x86/include/asm/amd_nb.h | 11 +++++++++++ arch/x86/kernel/amd_nb.c | 4 +--- arch/x86/kernel/aperture_64.c | 4 ++-- 3 files changed, 14 insertions(+), 5 deletions(-)
arch/x86/include/asm/amd_nb.h +11 -0
@@ -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
arch/x86/kernel/amd_nb.c +1 -3
@@ -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 ;/*
arch/x86/kernel/aperture_64.c +2 -2
@@ -262,7 +262,7 @@ 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 (!early_pci_allowed ()) +if (!early_pci_allowed () || !amd_gart_present ())return ;/* This is mostly duplicate of iommu_hole_init */
@@ -356,7 +356,7 @@ int __init gart_iommu_hole_init(void)int i, node;if (gart_iommu_aperture_disabled || !fix_aperture || - !early_pci_allowed ()) + !early_pci_allowed () || !amd_gart_present ())return -ENODEV ;pr_info ("Checking aperture...\n" );
Signature · 8 lines
2.3.3
--
Regards/Gruss,
Boris.
ECO tip #101: Trim your mails when you reply.
--Quoted 19 lines · Borislav Petkov
Quoted 1 line · Aravind Gopalakrishnan
Quoted 1 line
No need - I can amend the local copy I have here.
Here's what I did:
---
From: Aravind Gopalakrishnan <[email protected]>
Date: Wed, 1 Apr 2015 09:32:08 -0500
Subject: [PATCH] x86/gart: Check for GART support before accessing GART
registers
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bitGART registers are not present in newer AMD processors (Fam15h, Model 10h and later). So, avoid accesses to GART registers in PCI config space by returning early in
early_gart_iommu_check()andgart_iommu_hole_init()if GART is not available.
Quoted 8 lines
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 a aperture memory hole
hah, someone should fix the typo in that message:
s/a a/a
Quoted 10 lines
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 withmemblock_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.Refactor the family check used in
amd_nbinto an inline function so we
s/amd_nb/amd_nb.c
arch/x86/include/asm/amd_nb.h static inline u16 amd_get_node_id(struct pci_dev *pdev) · line 103
Quoted 4 lines
arch/x86/include/asm/amd_nb.h +3
-0+static inline bool amd_gart_present (void ) +{ +/* GART present only on Fam15h upto model 0fh */
s/h u/h, u
Quoted 18 lines
arch/x86/kernel/amd_nb.c +1
-0+if (amd_gart_present ()) amd_northbridges.flags |=AMD_NB_GART ;/* arch/x86/kernel/aperture_64.c +1
-1@@ -262,7 +262,7 @@ 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 (!early_pci_allowed ()) +if (!early_pci_allowed () || !amd_gart_present ())return ;/* This is mostly duplicate of iommu_hole_init */
arch/x86/kernel/aperture_64.c int __init gart_iommu_hole_init(void) · line 360
Quoted 6 lines
arch/x86/kernel/aperture_64.c +1
-1@@ -356,7 +356,7 @@ int __init gart_iommu_hole_init(void)int i, node;if (gart_iommu_aperture_disabled || !fix_aperture || - !early_pci_allowed ()) + !early_pci_allowed () || !amd_gart_present ())return -ENODEV ;
So what happens if !early_pci_allowed() but the GART is present? We'll set amd_northbridges.flags |= AMD_NB_GART, but won't run any of the setup code in aperture_64.c, right? Is that a valid setup?
Thanks,
Ingo
Quoted 21 lines · Ingo Molnar
Quoted 21 lines
arch/x86/kernel/amd_nb.c +1
-0+if (amd_gart_present ()) amd_northbridges.flags |=AMD_NB_GART ;/* arch/x86/kernel/aperture_64.c +1
-1@@ -262,7 +262,7 @@ 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 (!early_pci_allowed ()) +if (!early_pci_allowed () || !amd_gart_present ())return ;/* This is mostly duplicate of iommu_hole_init */ @@ -356,7 +356,7 @@ int __init gart_iommu_hole_init(void)int i, node;
Quoted 8 lines
arch/x86/kernel/aperture_64.c int __init gart_iommu_hole_init(void) · line 360
Quoted 4 lines
arch/x86/kernel/aperture_64.c +1
-1if (gart_iommu_aperture_disabled || !fix_aperture || - !early_pci_allowed ()) + !early_pci_allowed () || !amd_gart_present ())return -ENODEV ;So what happens if !early_pci_allowed() but the GART is present? We'll set amd_northbridges.flags |=
AMD_NB_GART, but won't run any of the setup code inaperture_64.c, right? Is that a valid setup?
It might be a valid setup. But it would work correctly only if BIOS did the right thing with setting up aperture space.
If !early_pci_allowed() and BIOS did not setup aperture correctly, that would have caused problems already. But it has not been an issue so far right?
Thanks,
-Aravind.
Quoted 10 lines · Aravind Gopalakrishnan
Quoted 3 lines
So what happens if !early_pci_allowed() but the GART is present? We'll set amd_northbridges.flags |=
AMD_NB_GART, but won't run any of the setup code inaperture_64.c, right? Is that a valid setup?It might be a valid setup. But it would work correctly only if BIOS did the right thing with setting up aperture space.
If !early_pci_allowed() and BIOS did not setup aperture correctly, that would have caused problems already.
But it has not been an issue so far right?
I think the right question to ask is:
What are we doing when there's no GART?
So what you probably want to do is:
if (!amd_gart_present()) return -ENODEV;
*before* the compound check.
Now, if we have detected a GART, i.e., amd_gart_present() is TRUE, the code will behave as it used to behave before.
This should be the least intrusive change going forward IMHO.
Signature · 5 lines
Regards/Gruss,
Boris.
ECO tip #101: Trim your mails when you reply.
--Quoted 19 lines · Borislav Petkov
Quoted 9 lines · Aravind Gopalakrishnan
Quoted 3 lines
So what happens if !early_pci_allowed() but the GART is present? We'll set amd_northbridges.flags |=
AMD_NB_GART, but won't run any of the setup code inaperture_64.c, right? Is that a valid setup?It might be a valid setup. But it would work correctly only if BIOS did the right thing with setting up aperture space.
If !early_pci_allowed() and BIOS did not setup aperture correctly, that would have caused problems already.
But it has not been an issue so far right?I think the right question to ask is:
What are we doing when there's no GART?
So what you probably want to do is:
if (!amd_gart_present()) return -ENODEV;
Quoted 7 lines
*before* the compound check.
Now, if we have detected a GART, i.e.,
amd_gart_present()is TRUE, the code will behave as it used to behave before.This should be the least intrusive change going forward IMHO.
Okay. I'll do that and correct the typos Ingo pointed out earlier and resend.
Thanks,
-Aravind.
Quoted 2 lines · Aravind Gopalakrishnan
Okay. I'll do that and correct the typos Ingo pointed out earlier and resend.
Btw, I think you should do the same in early_gart_iommu_check() too.
Doing the testing this way would mean that we first are testing for GART hw presence and then do the rest of checks. If no GART hw, the rest of the checks are meaningless.
Also, when testing do a "pci=noearly" boot which should make
!early_pci_allowed()
true and thus test that path too.
Thanks.
Signature · 5 lines
Regards/Gruss,
Boris.
ECO tip #101: Trim your mails when you reply.
--Quoted 9 lines · Borislav Petkov
Quoted 2 lines · Aravind Gopalakrishnan
Okay. I'll do that and correct the typos Ingo pointed out earlier and resend.
Btw, I think you should do the same in
early_gart_iommu_check()too.Doing the testing this way would mean that we first are testing for GART hw presence and then do the rest of checks. If no GART hw, the rest of the checks are meaningless.
Quoted 6 lines
Also, when testing do a "pci=noearly" boot which should make
!early_pci_allowed()
true and thus test that path too.
Here are results from further testing:
1. on platforms with both iommu and gart
- with pci=noearly, we break out of init routines in aperture_64.c
early. amd_iommu_init() will run through it's init routine.
- if amd_iommu_init() fails somewhere, we fall back togart_iommu_init()
- gart_iommu_init() fails since gart_iommu_aperture is not set.
- fall back to swiotlb.
- with amd_iommu=off
- init routines in aperture_64.c run fine as bothamd_gart_present() and early_pci_allowed() are true
- amd_iommu_detect() fails due to command line arg.
- fall back to gart iommu
- with pci=noearly and amd_iommu=off
- break out of aperture_64.c init routines, andamd_iommu_detect() fails.
- fall back to swiotlb
2. on platform with no gart but iommu present,
- pci=noearly option is not relevant as we break before that due to
!amd_gart_present()
- if amd_iommu_init() fails somewhere, we fall back to swiotlb,
else use iommu
3. on platforms with no gart and no iommu
- use swiotlb regardless of any command line options passed
Thanks,
-Aravind.
Quoted 35 lines · Aravind Gopalakrishnan
Quoted 9 lines · Borislav Petkov
Quoted 2 lines · Aravind Gopalakrishnan
Okay. I'll do that and correct the typos Ingo pointed out earlier and resend.
Btw, I think you should do the same in
early_gart_iommu_check()too.Doing the testing this way would mean that we first are testing for GART hw presence and then do the rest of checks. If no GART hw, the rest of the checks are meaningless.
Quoted 6 lines
Also, when testing do a "pci=noearly" boot which should make
!early_pci_allowed()
true and thus test that path too.
Here are results from further testing:
1. on platforms with both iommu and gart
- with pci=noearly, we break out of init routines inaperture_64.c
early.amd_iommu_init()will run through it's init routine.
- ifamd_iommu_init()fails somewhere, we fall back togart_iommu_init()
-gart_iommu_init()fails sincegart_iommu_apertureis not set.
- fall back to swiotlb.
- with amd_iommu=off
- init routines inaperture_64.crun fine as bothamd_gart_present()
andearly_pci_allowed()are true
-amd_iommu_detect()fails due to command line arg.
- fall back to gart iommu
- with pci=noearly and amd_iommu=off
- break out ofaperture_64.cinit routines, andamd_iommu_detect()
fails.
- fall back to swiotlb
Quoted 8 lines
2. on platform with no gart but iommu present,
- pci=noearly option is not relevant as we break before that due to
!amd_gart_present()
- ifamd_iommu_init()fails somewhere, we fall back to swiotlb, else use
iommu3. on platforms with no gart and no iommu
- use swiotlb regardless of any command line options passed
I don't see anything out of the ordinary but then again I don't know this code so...
@joro, can you please double-check?
Thanks.
Signature · 5 lines
Regards/Gruss,
Boris.
ECO tip #101: Trim your mails when you reply.
--Hi,
tried this patch on my HP Pavilion notebook wiht AMD Phenom II processor. The BIOS has no IOMMU option.
With this patch I'm still seeing the AGP messages mention in the patch description. Is this processor affected at all? /proc/cpuinfo says "cpu family" 16. That is 10h?
Regards,
jvp.
Borislav Petkov wrote on 04/02/2015 19:17:
Quoted 31 lines · Aravind Gopalakrishnan
Quoted 1 line
Quoted 1 line
No need - I can amend the local copy I have here.
Here's what I did:
---
From: Aravind Gopalakrishnan <[email protected]>
Date: Wed, 1 Apr 2015 09:32:08 -0500
Subject: [PATCH] x86/gart: Check for GART support before accessing GART
registers
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bitGART registers are not present in newer AMD processors (Fam15h, Model 10h and later). So, avoid accesses to GART registers in PCI config space by returning early in
early_gart_iommu_check()andgart_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 a 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]
<snip>
Borislav Petkov
maintainerQuoted 1 line · Jörg-Volker Peetz
Hi,
Hi,
please do not top-post.
Quoted 4 lines
tried this patch on my HP Pavilion notebook wiht AMD Phenom II processor. The BIOS has no IOMMU option.
With this patch I'm still seeing the AGP messages mention in the patch description. Is this processor affected at all?
Your CPU has an AGP bridge but the BIOS doesn't leave an aperture so the kernel has to do it. Read the message.
Quoted 1 line
/proc/cpuinfosays "cpu family" 16. That is 10h?
Yes.
Signature · 5 lines
Regards/Gruss,
Boris.
ECO tip #101: Trim your mails when you reply.
--Hi,
Borislav Petkov wrote on 04/14/2015 13:33:
Quoted 12 lines · Jörg-Volker Peetz
Quoted 1 line
Hi,
Hi,
please do not top-post.
Quoted 4 lines
tried this patch on my HP Pavilion notebook wiht AMD Phenom II processor. The BIOS has no IOMMU option.
With this patch I'm still seeing the AGP messages mention in the patch description. Is this processor affected at all?
Quoted 7 lines
Your CPU has an AGP bridge but the BIOS doesn't leave an aperture so the kernel has to do it. Read the message.
Quoted 1 line
/proc/cpuinfosays "cpu family" 16. That is 10h?Yes.
thank you for the explanation. Somehow I thought I had a newer CPU w/o GART registers, but I misread the patch. Maybe wishful thinking, in the hope to save 64 MB :-(
Sorry for the noise.
Signature · 2 lines
Regards, jvp.
Borislav Petkov
maintainerQuoted 3 lines · Jörg-Volker Peetz
thank you for the explanation. Somehow I thought I had a newer CPU w/o GART registers, but I misread the patch. Maybe wishful thinking, in the hope to save 64 MB :-(
64MB out of how much? Does it even matter?
You could turn off CONFIG_GART_IOMMU and not build aperture_64.c at all.
I think then you'll fall back to swiotlb or whatever and that will most likely slow down DMA ... the story was something like that, at least, memory's hazy. Joerg's on CC, he'll correct me if I'm talking bull.
Signature · 5 lines
Regards/Gruss,
Boris.
ECO tip #101: Trim your mails when you reply.
--Quoted 7 lines · Borislav Petkov
64MB out of how much? Does it even matter?
You could turn off
CONFIG_GART_IOMMUand not buildaperture_64.cat all.I think then you'll fall back to swiotlb or whatever and that will most likely slow down DMA ... the story was something like that, at least, memory's hazy. Joerg's on CC, he'll correct me if I'm talking bull.
Using SWIOTLB will also eat 64MB of RAM, so it doesn't matter if you use it instead of the GART driver. So whether your machine has GART or not, it will cost you 64MB of RAM for DMA remapping (unless you have a real IOMMU).
Joerg