Re: [tip:x86/urgent] x86/Sandy Bridge: reserve pages when integrated graphics is present

Re: [tip:x86/urgent] x86/Sandy Bridge: reserve pages when integrated graphics is present

[Jesse Barnes]

On Fri, 11 Jan 2013 14:39:04 -0800
tip-bot for Jesse Barnes <jbarnes@virtuousgeek.org> wrote:

> Commit-ID:  a9acc5365dbda29f7be2884efb63771dc24bd815
> Gitweb:     http://git.kernel.org/tip/a9acc5365dbda29f7be2884efb63771dc24bd815
> Author:     Jesse Barnes <jbarnes@virtuousgeek.org>
> AuthorDate: Wed, 14 Nov 2012 20:43:31 +0000
> Committer:  H. Peter Anvin <hpa@linux.intel.com>
> CommitDate: Fri, 11 Jan 2013 14:26:38 -0800
> 
> x86/Sandy Bridge: reserve pages when integrated graphics is present
> 
> SNB graphics devices have a bug that prevent them from accessing certain
> memory ranges, namely anything below 1M and in the pages listed in the
> table.  So reserve those at boot if set detect a SNB gfx device on the
> CPU to avoid GPU hangs.
> 
> Stephane Marchesin had a similar patch to the page allocator awhile
> back, but rather than reserving pages up front, it leaked them at
> allocation time.
> 
> [ hpa: made a number of stylistic changes, marked arrays as static
>   const, and made less verbose; use "memblock=debug" for full
>   verbosity. ]
> 
> Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
> ---
>  arch/x86/kernel/setup.c | 78 +++++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 78 insertions(+)
> 
> diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
> index 23ddd55..9dcb325 100644
> --- a/arch/x86/kernel/setup.c
> +++ b/arch/x86/kernel/setup.c
> @@ -610,6 +610,81 @@ static __init void reserve_ibft_region(void)
>  
>  static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
>  
> +static bool __init snb_gfx_workaround_needed(void)
> +{
> +	int i;
> +	u16 vendor, devid;
> +	static const u16 snb_ids[] = {
> +		0x0102,
> +		0x0112,
> +		0x0122,
> +		0x0106,
> +		0x0116,
> +		0x0126,
> +		0x010a,
> +	};
> +
> +	/* Assume no if something weird is going on with PCI */
> +	if (!early_pci_allowed())
> +		return false;
> +
> +	vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
> +	if (vendor != 0x8086)
> +		return false;
> +
> +	devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
> +	for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
> +		if (devid == snb_ids[i])
> +			return true;
> +
> +	return false;
> +}
> +
> +/*
> + * Sandy Bridge graphics has trouble with certain ranges, exclude
> + * them from allocation.
> + */
> +static void __init trim_snb_memory(void)
> +{
> +	static const unsigned long bad_pages[] = {
> +		0x20050000,
> +		0x20110000,
> +		0x20130000,
> +		0x20138000,
> +		0x40004000,
> +	};
> +	int i;
> +
> +	if (!snb_gfx_workaround_needed())
> +		return;
> +
> +	printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
> +
> +	/*
> +	 * Reserve all memory below the 1 MB mark that has not
> +	 * already been reserved.
> +	 */
> +	memblock_reserve(0, 1<<20);
> +	
> +	for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
> +		if (memblock_reserve(bad_pages[i], PAGE_SIZE))
> +			printk(KERN_WARNING "failed to reserve 0x%08lx\n",
> +			       bad_pages[i]);
> +	}
> +}
> +
> +/*
> + * Here we put platform-specific memory range workarounds, i.e.
> + * memory known to be corrupt or otherwise in need to be reserved on
> + * specific platforms.
> + *
> + * If this gets used more widely it could use a real dispatch mechanism.
> + */
> +static void __init trim_platform_memory_ranges(void)
> +{
> +	trim_snb_memory();
> +}
> +
>  static void __init trim_bios_range(void)
>  {
>  	/*
> @@ -630,6 +705,7 @@ static void __init trim_bios_range(void)
>  	 * take them out.
>  	 */
>  	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
> +
>  	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
>  }
>  
> @@ -908,6 +984,8 @@ void __init setup_arch(char **cmdline_p)
>  
>  	setup_real_mode();
>  
> +	trim_platform_memory_ranges();
> +
>  	init_gbpages();
>  
>  	/* max_pfn_mapped is updated here */
> 

Great, thanks Peter.

Stephane, care to try this out?

-- 
Jesse Barnes, Intel Open Source Technology Center

Re: [tip:x86/urgent] x86/Sandy Bridge: reserve pages when integrated graphics is present

[H. Peter Anvin]

On 01/11/2013 03:00 PM, Jesse Barnes wrote:
> 
> Great, thanks Peter.
> 
> Stephane, care to try this out?
> 

For the record, I can confirm that it reserves the appropriate ranges on
my personal Sandy Bridge system; I don't have any test which can provoke
the use of these pages, so I can't say I have "fully tested" this.

	-hpa

Re: [tip:x86/urgent] x86/Sandy Bridge: reserve pages when integrated graphics is present

[Mathias Krause]

Hi Peter,

Am 12.01.2013 um 00:00 schrieb Jesse Barnes <jbarnes@virtuousgeek.org>:

> On Fri, 11 Jan 2013 14:39:04 -0800
> tip-bot for Jesse Barnes <jbarnes@virtuousgeek.org> wrote:
> 
>> Commit-ID:  a9acc5365dbda29f7be2884efb63771dc24bd815
>> Gitweb:     http://git.kernel.org/tip/a9acc5365dbda29f7be2884efb63771dc24bd815
>> Author:     Jesse Barnes <jbarnes@virtuousgeek.org>
>> AuthorDate: Wed, 14 Nov 2012 20:43:31 +0000
>> Committer:  H. Peter Anvin <hpa@linux.intel.com>
>> CommitDate: Fri, 11 Jan 2013 14:26:38 -0800
>> 
>> x86/Sandy Bridge: reserve pages when integrated graphics is present
>> 
>> SNB graphics devices have a bug that prevent them from accessing certain
>> memory ranges, namely anything below 1M and in the pages listed in the
>> table.  So reserve those at boot if set detect a SNB gfx device on the
>> CPU to avoid GPU hangs.
>> 
>> Stephane Marchesin had a similar patch to the page allocator awhile
>> back, but rather than reserving pages up front, it leaked them at
>> allocation time.
>> 
>> [ hpa: made a number of stylistic changes, marked arrays as static
>>  const, and made less verbose; use "memblock=debug" for full
>>  verbosity. ]

That stylistic change introduced the unwanted side effect of leaving the tables in memory after __init cleanup. But that can be fixed, see below :)

>> 
>> Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
>> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
>> ---
>> arch/x86/kernel/setup.c | 78 +++++++++++++++++++++++++++++++++++++++++++++++++
>> 1 file changed, 78 insertions(+)
>> 
>> diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
>> index 23ddd55..9dcb325 100644
>> --- a/arch/x86/kernel/setup.c
>> +++ b/arch/x86/kernel/setup.c
>> @@ -610,6 +610,81 @@ static __init void reserve_ibft_region(void)
>> 
>> static unsigned reserve_low = CONFIG_X86_RESERVE_LOW << 10;
>> 
>> +static bool __init snb_gfx_workaround_needed(void)
>> +{
>> +	int i;
>> +	u16 vendor, devid;
>> +	static const u16 snb_ids[] = {

__initconst is missing here

>> +		0x0102,
>> +		0x0112,
>> +		0x0122,
>> +		0x0106,
>> +		0x0116,
>> +		0x0126,
>> +		0x010a,
>> +	};
>> +
>> +	/* Assume no if something weird is going on with PCI */
>> +	if (!early_pci_allowed())
>> +		return false;
>> +
>> +	vendor = read_pci_config_16(0, 2, 0, PCI_VENDOR_ID);
>> +	if (vendor != 0x8086)
>> +		return false;
>> +
>> +	devid = read_pci_config_16(0, 2, 0, PCI_DEVICE_ID);
>> +	for (i = 0; i < ARRAY_SIZE(snb_ids); i++)
>> +		if (devid == snb_ids[i])
>> +			return true;
>> +
>> +	return false;
>> +}
>> +
>> +/*
>> + * Sandy Bridge graphics has trouble with certain ranges, exclude
>> + * them from allocation.
>> + */
>> +static void __init trim_snb_memory(void)
>> +{
>> +	static const unsigned long bad_pages[] = {

here, too

>> +		0x20050000,
>> +		0x20110000,
>> +		0x20130000,
>> +		0x20138000,
>> +		0x40004000,
>> +	};
>> +	int i;
>> +
>> +	if (!snb_gfx_workaround_needed())
>> +		return;
>> +
>> +	printk(KERN_DEBUG "reserving inaccessible SNB gfx pages\n");
>> +
>> +	/*
>> +	 * Reserve all memory below the 1 MB mark that has not
>> +	 * already been reserved.
>> +	 */
>> +	memblock_reserve(0, 1<<20);
>> +	
>> +	for (i = 0; i < ARRAY_SIZE(bad_pages); i++) {
>> +		if (memblock_reserve(bad_pages[i], PAGE_SIZE))
>> +			printk(KERN_WARNING "failed to reserve 0x%08lx\n",
>> +			       bad_pages[i]);
>> +	}
>> +}
>> +
>> +/*
>> + * Here we put platform-specific memory range workarounds, i.e.
>> + * memory known to be corrupt or otherwise in need to be reserved on
>> + * specific platforms.
>> + *
>> + * If this gets used more widely it could use a real dispatch mechanism.
>> + */
>> +static void __init trim_platform_memory_ranges(void)
>> +{
>> +	trim_snb_memory();
>> +}
>> +
>> static void __init trim_bios_range(void)
>> {
>> 	/*
>> @@ -630,6 +705,7 @@ static void __init trim_bios_range(void)
>> 	 * take them out.
>> 	 */
>> 	e820_remove_range(BIOS_BEGIN, BIOS_END - BIOS_BEGIN, E820_RAM, 1);
>> +
>> 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
>> }
>> 
>> @@ -908,6 +984,8 @@ void __init setup_arch(char **cmdline_p)
>> 
>> 	setup_real_mode();
>> 
>> +	trim_platform_memory_ranges();
>> +
>> 	init_gbpages();
>> 
>> 	/* max_pfn_mapped is updated here */
>> 


Regards,
Mathias

[tip:x86/urgent] x86/Sandy Bridge: mark arrays in __init functions as __initconst

[tip-bot for H. Peter Anvin]

Commit-ID:  ab3cd8670e0b3fcde7f029e1503ed3c5138e9571
Gitweb:     http://git.kernel.org/tip/ab3cd8670e0b3fcde7f029e1503ed3c5138e9571
Author:     H. Peter Anvin <hpa@linux.intel.com>
AuthorDate: Sun, 13 Jan 2013 20:36:39 -0800
Committer:  H. Peter Anvin <hpa@linux.intel.com>
CommitDate: Sun, 13 Jan 2013 20:36:39 -0800

x86/Sandy Bridge: mark arrays in __init functions as __initconst

Mark static arrays as __initconst so they get removed when the init
sections are flushed.

Reported-by: Mathias Krause <minipli@googlemail.com>
Link: http://lkml.kernel.org/r/75F4BEE6-CB0E-4426-B40B-697451677738@googlemail.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
---
 arch/x86/kernel/setup.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index 9dcb325..18182d1 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -614,7 +614,7 @@ static bool __init snb_gfx_workaround_needed(void)
 {
 	int i;
 	u16 vendor, devid;
-	static const u16 snb_ids[] = {
+	static const __initconst u16 snb_ids[] = {
 		0x0102,
 		0x0112,
 		0x0122,
@@ -646,7 +646,7 @@ static bool __init snb_gfx_workaround_needed(void)
  */
 static void __init trim_snb_memory(void)
 {
-	static const unsigned long bad_pages[] = {
+	static const __initconst unsigned long bad_pages[] = {
 		0x20050000,
 		0x20110000,
 		0x20130000,