[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <35ba181f-c182-aa74-07e2-fcff94bf345a@amd.com>
Date: Fri, 3 Dec 2021 11:28:58 -0600
From: Tom Lendacky <thomas.lendacky@....com>
To: Lucas De Marchi <lucas.demarchi@...el.com>,
Joerg Roedel <joro@...tes.org>
Cc: x86@...nel.org, Dave Hansen <dave.hansen@...ux.intel.com>,
Borislav Petkov <bp@...en8.de>, Ingo Molnar <mingo@...hat.com>,
Thomas Gleixner <tglx@...utronix.de>, hpa@...or.com,
Andy Lutomirski <luto@...nel.org>,
Peter Zijlstra <peterz@...radead.org>,
Brijesh Singh <brijesh.singh@....com>,
linux-kernel@...r.kernel.org, Joerg Roedel <jroedel@...e.de>
Subject: Re: [PATCH] x86/mm: Fix PAGE_KERNEL_IO removal breakage
On 12/2/21 6:25 PM, Lucas De Marchi wrote:
> On Thu, Dec 02, 2021 at 07:55:14AM -0800, Lucas De Marchi wrote:
>> On Thu, Dec 02, 2021 at 03:46:46PM +0100, Joerg Roedel wrote:
>>> From: Joerg Roedel <jroedel@...e.de>
>>>
>>> The removal of PAGE_KERNEL_IO broke SEV-ES because it changed the
>>> mapping of ioremap and some fixmap areas (like the local APIC page)
>>> from unencrypted to encrypted. Change those mappings back to
>>> be unencrypted.
>>>
>>> Cc: Lucas De Marchi <lucas.demarchi@...el.com>
>>> Fixes: 27dff0f58bde ("x86/mm: Nuke PAGE_KERNEL_IO")
>>> Signed-off-by: Joerg Roedel <jroedel@...e.de>
>>
>> oops, missed the fact PAGE_KERNEL had `| ENC` while PAGE_KERENL_IO
>> didn't have it. Thanks for the fixup.
>
> on a second thought, the fact that PAGE_KERNEL is _not_ the same as
> PAGE_KERNEL_IO, completely invalidates those 2 patches I sent. It seems
> I screwed it up big here.
>
> About the first patch,
> 6b2a2138cf36 ("drm/i915/gem: Stop using PAGE_KERNEL_IO"),
> I didn't notice any regression on the i915
> side though. Is it safe to keep it? Otherwise we are probably better
> off reverting everything.
>
> I'm wondering why the addition of memory encryption
> in 21729f81ce8a ("x86/mm: Provide general kernel support for memory
> encryption")
> didn't break io_mapping_init_wc() though as it had already done a
> s/PAGE_KERNEL_IO/PAGE_KERNEL/ in commit
> ac96b5566926 ("io-mapping.h: s/PAGE_KERNEL_IO/PAGE_KERNEL/")
If I follow it correctly, since SME/SEV-ES are X86_64 only,
io_mapping_init_wc() takes the ioremap_wc() path which uses
PAGE_KERNEL_IO. iomap_create_wc() is only called when
CONFIG_HAVE_ATOMIC_IOMAP is set, which isn't set for X86_64.
Thanks,
Tom
>
> thanks
> Lucas De Marchi
>
>>
>> Reviewed-by: Lucas De Marchi <lucas.demarchi@...el.com>
>>
>> Lucas De Marchi
>>
>>> ---
>>> arch/x86/include/asm/fixmap.h | 2 +-
>>> arch/x86/include/asm/pgtable_types.h | 21 +++++++++++----------
>>> arch/x86/mm/ioremap.c | 2 +-
>>> 3 files changed, 13 insertions(+), 12 deletions(-)
>>>
>>> diff --git a/arch/x86/include/asm/fixmap.h b/arch/x86/include/asm/fixmap.h
>>> index 5e186a69db10..a2eaf265f784 100644
>>> --- a/arch/x86/include/asm/fixmap.h
>>> +++ b/arch/x86/include/asm/fixmap.h
>>> @@ -173,7 +173,7 @@ static inline void __set_fixmap(enum
>>> fixed_addresses idx,
>>> * supported for MMIO addresses, so make sure that the memory encryption
>>> * mask is not part of the page attributes.
>>> */
>>> -#define FIXMAP_PAGE_NOCACHE PAGE_KERNEL_NOCACHE
>>> +#define FIXMAP_PAGE_NOCACHE PAGE_KERNEL_NOCACHE_NOENC
>>>
>>> /*
>>> * Early memremap routines used for in-place encryption. The mappings
>>> created
>>> diff --git a/arch/x86/include/asm/pgtable_types.h
>>> b/arch/x86/include/asm/pgtable_types.h
>>> index a87224767ff3..fc9b6995cb22 100644
>>> --- a/arch/x86/include/asm/pgtable_types.h
>>> +++ b/arch/x86/include/asm/pgtable_types.h
>>> @@ -208,16 +208,17 @@ enum page_cache_mode {
>>>
>>> #define __pgprot_mask(x) __pgprot((x) & __default_kernel_pte_mask)
>>>
>>> -#define PAGE_KERNEL __pgprot_mask(__PAGE_KERNEL | _ENC)
>>> -#define PAGE_KERNEL_NOENC __pgprot_mask(__PAGE_KERNEL
>>> | 0)
>>> -#define PAGE_KERNEL_RO __pgprot_mask(__PAGE_KERNEL_RO |
>>> _ENC)
>>> -#define PAGE_KERNEL_EXEC __pgprot_mask(__PAGE_KERNEL_EXEC |
>>> _ENC)
>>> -#define PAGE_KERNEL_EXEC_NOENC
>>> __pgprot_mask(__PAGE_KERNEL_EXEC | 0)
>>> -#define PAGE_KERNEL_ROX __pgprot_mask(__PAGE_KERNEL_ROX
>>> | _ENC)
>>> -#define PAGE_KERNEL_NOCACHE __pgprot_mask(__PAGE_KERNEL_NOCACHE
>>> | _ENC)
>>> -#define PAGE_KERNEL_LARGE __pgprot_mask(__PAGE_KERNEL_LARGE |
>>> _ENC)
>>> -#define PAGE_KERNEL_LARGE_EXEC
>>> __pgprot_mask(__PAGE_KERNEL_LARGE_EXEC | _ENC)
>>> -#define PAGE_KERNEL_VVAR __pgprot_mask(__PAGE_KERNEL_VVAR |
>>> _ENC)
>>> +#define PAGE_KERNEL __pgprot_mask(__PAGE_KERNEL
>>> | _ENC)
>>> +#define PAGE_KERNEL_NOENC
>>> __pgprot_mask(__PAGE_KERNEL | 0)
>>> +#define PAGE_KERNEL_RO
>>> __pgprot_mask(__PAGE_KERNEL_RO | _ENC)
>>> +#define PAGE_KERNEL_EXEC __pgprot_mask(__PAGE_KERNEL_EXEC
>>> | _ENC)
>>> +#define PAGE_KERNEL_EXEC_NOENC
>>> __pgprot_mask(__PAGE_KERNEL_EXEC | 0)
>>> +#define PAGE_KERNEL_ROX
>>> __pgprot_mask(__PAGE_KERNEL_ROX | _ENC)
>>> +#define PAGE_KERNEL_NOCACHE
>>> __pgprot_mask(__PAGE_KERNEL_NOCACHE | _ENC)
>>> +#define PAGE_KERNEL_NOCACHE_NOENC
>>> __pgprot_mask(__PAGE_KERNEL_NOCACHE | 0)
>>> +#define PAGE_KERNEL_LARGE
>>> __pgprot_mask(__PAGE_KERNEL_LARGE | _ENC)
>>> +#define PAGE_KERNEL_LARGE_EXEC
>>> __pgprot_mask(__PAGE_KERNEL_LARGE_EXEC | _ENC)
>>> +#define PAGE_KERNEL_VVAR __pgprot_mask(__PAGE_KERNEL_VVAR
>>> | _ENC)
>>>
>>> #endif /* __ASSEMBLY__ */
>>>
>>> diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
>>> index 3102dda4b152..4fe8d43d53bb 100644
>>> --- a/arch/x86/mm/ioremap.c
>>> +++ b/arch/x86/mm/ioremap.c
>>> @@ -243,7 +243,7 @@ __ioremap_caller(resource_size_t phys_addr,
>>> unsigned long size,
>>> * make sure the memory encryption attribute is enabled in the
>>> * resulting mapping.
>>> */
>>> - prot = PAGE_KERNEL;
>>> + prot = PAGE_KERNEL_NOENC;
>>> if ((io_desc.flags & IORES_MAP_ENCRYPTED) || encrypted)
>>> prot = pgprot_encrypted(prot);
>>>
>>> --
>>> 2.34.0
>>>
Powered by blists - more mailing lists