| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <55743353-60fb-4ad2-a0c2-6d5a3a2f8ac0@arm.com>
Date: Wed, 12 Feb 2025 18:48:40 +0000
From: Suzuki K Poulose <suzuki.poulose@....com>
To: Robin Murphy <robin.murphy@....com>, will@...nel.org, maz@...nel.org,
catalin.marinas@....com
Cc: linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
gregkh@...uxfoundation.org, aneesh.kumar@...nel.org, steven.price@....com,
Jean-Philippe Brucker <jean-philippe@...aro.org>,
Christoph Hellwig <hch@....de>, Tom Lendacky <thomas.lendacky@....com>
Subject: Re: [PATCH 1/1] arm64: realm: Use aliased addresses for device DMA to
shared buffers
Hi Robin
On 12/02/2025 18:18, Robin Murphy wrote:
> On 2025-02-12 5:14 pm, Suzuki K Poulose wrote:
>> When a device performs DMA to a shared buffer using physical addresses,
>> (without Stage1 translation), the device must use the "{I}PA address"
>> with the
>> top bit set in Realm. This is to make sure that a trusted device will
>> be able
>> to write to shared buffers as well as the protected buffers. Thus, a
>> Realm must
>> always program the full address including the "protection" bit, like
>> AMD SME
>> encryption bits.
>>
>> Add the support for this by providing arm64 version of the
>> phys_to_dma(). We
>> cannot use the __sme_mask as it assumes the "encryption" always "sets
>> a bit",
>> which is the opposite for CCA. i.e., "set a bit" for "decrypted"
>> address. So,
>> move the common code that can be reused by all - i.e., add
>> __phys_to_dma() and
>> __dma_to_phys() - and do the arch specific processing.
>>
>> Please note that the VMM needs to similarly make sure that the SMMU
>> Stage2 in
>> the Non-secure world is setup accordingly to map IPA at the
>> unprotected alias.
>>
>> Cc: Will Deacon <will@...nel.org>
>> Cc: Jean-Philippe Brucker <jean-philippe@...aro.org>
>> Cc: Catalin Marinas <catalin.marinas@....com>
>> Cc: Robin Murphy <robin.murphy@....com>
>> Cc: Steven Price <steven.price@....com>
>> Cc: Christoph Hellwig <hch@....de>
>> Cc: Tom Lendacky <thomas.lendacky@....com>
>> Signed-off-by: Suzuki K Poulose <suzuki.poulose@....com>
>> ---
>> arch/arm64/Kconfig | 1 +
>> arch/arm64/include/asm/dma-direct.h | 38 +++++++++++++++++++++++++++++
>> include/linux/dma-direct.h | 35 +++++++++++++++++---------
>> 3 files changed, 62 insertions(+), 12 deletions(-)
>> create mode 100644 arch/arm64/include/asm/dma-direct.h
>>
>> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
>> index fcdd0ed3eca8..7befe04106de 100644
>> --- a/arch/arm64/Kconfig
>> +++ b/arch/arm64/Kconfig
>> @@ -41,6 +41,7 @@ config ARM64
>> select ARCH_HAS_NMI_SAFE_THIS_CPU_OPS
>> select ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
>> select ARCH_HAS_NONLEAF_PMD_YOUNG if ARM64_HAFT
>> + select ARCH_HAS_PHYS_TO_DMA
>> select ARCH_HAS_PTE_DEVMAP
>> select ARCH_HAS_PTE_SPECIAL
>> select ARCH_HAS_HW_PTE_YOUNG
>> diff --git a/arch/arm64/include/asm/dma-direct.h b/arch/arm64/include/
>> asm/dma-direct.h
>> new file mode 100644
>> index 000000000000..37c3270542b8
>> --- /dev/null
>> +++ b/arch/arm64/include/asm/dma-direct.h
>> @@ -0,0 +1,38 @@
>> +/* SPDX-License-Identifier: GPL-2.0-only */
>> +#ifndef __ASM_DMA_DIRECT_H
>> +#define __ASM_DMA_DIRECT_H
>> +
>> +#include <asm/pgtable-prot.h>
>> +
>> +static inline unsigned long addr_to_shared(unsigned long addr)
>> +{
>> + if (is_realm_world())
>> + addr |= prot_ns_shared;
>> + return addr;
>> +}
>> +
>> +static inline unsigned long addr_to_private(unsigned long addr)
>> +{
>> + if (is_realm_world())
>> + addr &= prot_ns_shared - 1;
>> + return addr;
>> +}
>> +
>> +static inline dma_addr_t phys_to_dma(struct device *dev, phys_addr_t
>> paddr)
>> +{
>> + return __phys_to_dma(dev, paddr);
>> +}
>> +
>> +static inline dma_addr_t phys_to_dma_unencrypted(struct device *dev,
>> + phys_addr_t paddr)
>> +{
>> + return addr_to_shared(__phys_to_dma(dev, paddr));
>> +}
>> +#define phys_to_dma_unencrypted phys_to_dma_unencrypted
>> +
>> +static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t
>> dma_addr)
>> +{
>> + return addr_to_private(__dma_to_phys(dev, dma_addr));
>> +}
>> +
>> +#endif /* __ASM_DMA_DIRECT_H */
>> diff --git a/include/linux/dma-direct.h b/include/linux/dma-direct.h
>> index d7e30d4f7503..3e9bf6ca640e 100644
>> --- a/include/linux/dma-direct.h
>> +++ b/include/linux/dma-direct.h
>> @@ -72,18 +72,36 @@ static inline dma_addr_t dma_range_map_max(const
>> struct bus_dma_region *map)
>> return ret;
>> }
>> +static inline dma_addr_t __phys_to_dma(struct device *dev,
>> phys_addr_t paddr)
>> +{
>> + if (dev->dma_range_map)
>> + return translate_phys_to_dma(dev, paddr);
>> + return paddr;
>> +}
>> +
>> +static inline phys_addr_t __dma_to_phys(struct device *dev,
>> dma_addr_t dma_addr)
>> +{
>> + phys_addr_t paddr;
>> +
>> + if (dev->dma_range_map)
>> + paddr = translate_dma_to_phys(dev, dma_addr);
>> + else
>> + paddr = dma_addr;
>> +
>> + return paddr;
>> +}
>> +
>> #ifdef CONFIG_ARCH_HAS_PHYS_TO_DMA
>> #include <asm/dma-direct.h>
>> #ifndef phys_to_dma_unencrypted
>> #define phys_to_dma_unencrypted phys_to_dma
>> #endif
>> #else
>> +
>> static inline dma_addr_t phys_to_dma_unencrypted(struct device *dev,
>> phys_addr_t paddr)
>> {
>> - if (dev->dma_range_map)
>> - return translate_phys_to_dma(dev, paddr);
>> - return paddr;
>> + return __phys_to_dma(dev, paddr);
>> }
>> /*
>> @@ -94,19 +112,12 @@ static inline dma_addr_t
>> phys_to_dma_unencrypted(struct device *dev,
>> */
>> static inline dma_addr_t phys_to_dma(struct device *dev, phys_addr_t
>> paddr)
>> {
>> - return __sme_set(phys_to_dma_unencrypted(dev, paddr));
>> + return __sme_set(__phys_to_dma(dev, paddr));
>
> Hmm, it really feels like we should generalise __sme_{set,clr} at this
> level, rather than drag in the entire ARCH_HAS_PHYS_TO_DMA override for
> the purposes of setting/clearing an address bit just because the
> "generic" mechanism for doing that is unashamedly AMD-specific.
I was planning to convert that to use ARCH_HAS_PHYS_TO_DMA ;-) too, in a
separate patch, to get rid of this "one approach" works for all hook.
>
> Maybe something like:
>
> #define dma_shared(x) sme_clr(x)
> #define dma_private(x) sme_set(x)
>
> for x86, then go from there?
But this looks sensible too, can switch to this approach.
Thanks for the feedback.
Suzuki
>
> Thanks,
> Robin.
>
>> }
>> static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t
>> dma_addr)
>> {
>> - phys_addr_t paddr;
>> -
>> - if (dev->dma_range_map)
>> - paddr = translate_dma_to_phys(dev, dma_addr);
>> - else
>> - paddr = dma_addr;
>> -
>> - return __sme_clr(paddr);
>> + return __sme_clr(__dma_to_phys(dev, dma_addr));
>> }
>> #endif /* !CONFIG_ARCH_HAS_PHYS_TO_DMA */
>
Powered by blists - more mailing lists