From: Jacob Pan <jacob.pan@linux.microsoft.com>
To: Shyam Saini <shyamsaini@linux.microsoft.com>
Cc: iommu@lists.linux.dev, linux-arm-kernel@lists.infradead.org,
devicetree@vger.kernel.org, virtualization@lists.linux.dev,
jgg@ziepe.ca, will@kernel.org, eric.auger@redhat.com,
code@tyhicks.com, eahariha@linux.microsoft.com,
vijayb@linux.microsoft.com, bboscaccy@linux.microsoft.com
Subject: Re: [PATCH v3 3/3] arm-smmu: select suitable MSI IOVA
Date: Mon, 11 Aug 2025 15:34:44 -0700 [thread overview]
Message-ID: <20250811153444.000071ab@linux.microsoft.com> (raw)
In-Reply-To: <20250806215539.1240561-4-shyamsaini@linux.microsoft.com>
On Wed, 6 Aug 2025 14:55:39 -0700
Shyam Saini <shyamsaini@linux.microsoft.com> wrote:
> Currently ARM SMMU drivers hardcode PCI MSI IOVA address.
> Not all the platform have same memory mappings and some platform
> could have this address already being mapped for something else.
> This can lead to collision and as a consequence the MSI IOVA addr
> range is never reserved.
>
> Fix this by reserving faulty IOVA range and selecting alternate
> MSI_IOVA suitable for the intended platform.
>
> Example of reserving faulty IOVA range for PCIE device in the DTS:
>
> reserved-memory {
> #address-cells = <2>;
> #size-cells = <2>;
> faulty_iova: resv_faulty {
> iommu-addresses = <&pcieX 0x0 0x8000000 0x0 0x100000>;
> };
> };
>
> &pcieX {
> memory-region = <&faulty_iova>;
> };
>
> Suggested-by: Jason Gunthorpe <jgg@ziepe.ca>
> Link: https://lore.kernel.org/linux-iommu/20250416181759.GF493866@ziepe.ca/
> Signed-off-by: Shyam Saini <shyamsaini@linux.microsoft.com>
> ---
> drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 27
> +++++++++++++++------ drivers/iommu/arm/arm-smmu/arm-smmu.c |
> 27 +++++++++++++++------ include/linux/iommu.h
> | 25 +++++++++++++++++++ 3 files changed, 65 insertions(+), 14
> deletions(-)
>
> diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c index
> 5f46140a8f3fa..0dfb522e173f0 100644 ---
> a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c +++
> b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c @@ -3642,17 +3642,30 @@
> static int arm_smmu_of_xlate(struct device *dev, static void
> arm_smmu_get_resv_regions(struct device *dev, struct list_head *head)
> {
> - struct iommu_resv_region *region;
> int prot = IOMMU_WRITE | IOMMU_NOEXEC | IOMMU_MMIO;
>
> - region = iommu_alloc_resv_region(MSI_IOVA_BASE,
> MSI_IOVA_LENGTH,
> - prot, IOMMU_RESV_SW_MSI,
> GFP_KERNEL);
> - if (!region)
> - return;
> -
> - list_add_tail(®ion->list, head);
> + static const u64 msi_bases[] = { MSI_IOVA_BASE,
> MSI_IOVA_BASE2 };
> iommu_dma_get_resv_regions(dev, head);
> +
> + /*
> + * Use the first msi_base that does not intersect with a
> platform
> + * reserved region. The SW MSI base selection is entirely
> arbitrary.
> + */
> + for (int i = 0; i != ARRAY_SIZE(msi_bases); i++) {
nit: i < ARRAY_SIZE(msi_bases) is more robust
> + struct iommu_resv_region *region;
> +
> + if (resv_region_intersects(msi_bases[i],
> MSI_IOVA_LENGTH, head))
> + continue;
> +
> + region = iommu_alloc_resv_region(msi_bases[i],
> MSI_IOVA_LENGTH, prot,
> + IOMMU_RESV_SW_MSI,
> GFP_KERNEL);
> + if (!region)
> + return;
> +
> + list_add_tail(®ion->list, head);
> + return;
> + }
Maybe add a warning if both msi_bases are conflicting. Or you can keep
searching until a non-intersecting region is found. e.g. let
resv_region_intersects() return a valid region.
> }
>
> /*
> diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c
> b/drivers/iommu/arm/arm-smmu/arm-smmu.c index
> 4a07650911991..84b74b8519386 100644 ---
> a/drivers/iommu/arm/arm-smmu/arm-smmu.c +++
> b/drivers/iommu/arm/arm-smmu/arm-smmu.c @@ -1600,17 +1600,30 @@
> static int arm_smmu_of_xlate(struct device *dev, static void
> arm_smmu_get_resv_regions(struct device *dev, struct list_head *head)
> {
> - struct iommu_resv_region *region;
> int prot = IOMMU_WRITE | IOMMU_NOEXEC | IOMMU_MMIO;
>
> - region = iommu_alloc_resv_region(MSI_IOVA_BASE,
> MSI_IOVA_LENGTH,
> - prot, IOMMU_RESV_SW_MSI,
> GFP_KERNEL);
> - if (!region)
> - return;
> -
> - list_add_tail(®ion->list, head);
> + static const u64 msi_bases[] = { MSI_IOVA_BASE,
> MSI_IOVA_BASE2 };
> iommu_dma_get_resv_regions(dev, head);
> +
> + /*
> + * Use the first msi_base that does not intersect with a
> platform
> + * reserved region. The SW MSI base selection is entirely
> arbitrary.
> + */
> + for (int i = 0; i != ARRAY_SIZE(msi_bases); i++) {
> + struct iommu_resv_region *region;
> +
> + if (resv_region_intersects(msi_bases[i],
> MSI_IOVA_LENGTH, head))
> + continue;
> +
> + region = iommu_alloc_resv_region(msi_bases[i],
> MSI_IOVA_LENGTH, prot,
> + IOMMU_RESV_SW_MSI,
> GFP_KERNEL);
> + if (!region)
> + return;
> +
> + list_add_tail(®ion->list, head);
> + return;
> + }
> }
>
> static int arm_smmu_def_domain_type(struct device *dev)
> diff --git a/include/linux/iommu.h b/include/linux/iommu.h
> index 517f908296436..6750ecdebaa94 100644
> --- a/include/linux/iommu.h
> +++ b/include/linux/iommu.h
> @@ -1509,14 +1509,39 @@ static inline void iommu_debugfs_setup(void)
> {}
> #ifdef CONFIG_IOMMU_DMA
> #define MSI_IOVA_BASE 0x8000000
> +#define MSI_IOVA_BASE2 0xA0000000
> #define MSI_IOVA_LENGTH 0x100000
>
> +static inline bool resv_region_intersects(phys_addr_t msi_base,
> size_t length,
> + struct list_head
> *resv_region_list) +{
> + struct iommu_resv_region *region;
> + phys_addr_t start, end, resv_region_end;
> +
> + start = msi_base;
> + end = start + length - 1;
> + list_for_each_entry(region, resv_region_list, list) {
> + resv_region_end = region->start + region->length - 1;
> + /* overlap detected */
> + if (!(start > resv_region_end || end <
> region->start))
> + return true;
> + }
> +
> + return false;
> +}
> +
> int iommu_get_msi_cookie(struct iommu_domain *domain, dma_addr_t
> base); #else /* CONFIG_IOMMU_DMA */
> static inline int iommu_get_msi_cookie(struct iommu_domain *domain,
> dma_addr_t base) {
> return -ENODEV;
> }
> +
> +static inline bool resv_region_intersects(phys_addr_t msi_base,
> size_t length,
> + struct list_head
> *resv_region_list) +{
> + return false;
> +}
> #endif /* CONFIG_IOMMU_DMA */
>
> /*
prev parent reply other threads:[~2025-08-11 22:37 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-08-06 21:55 [PATCH v3 0/3] arm-smmu: select suitable MSI IOVA Shyam Saini
2025-08-06 21:55 ` [PATCH v3 1/3] arm-smmu: move MSI_IOVA macro definitions Shyam Saini
2025-08-07 14:32 ` kernel test robot
2025-08-06 21:55 ` [PATCH v3 2/3] iommu/of: fix device tree configuration for PCI devices Shyam Saini
2025-08-08 12:15 ` Jason Gunthorpe
2025-08-14 23:30 ` Shyam Saini
2025-08-15 0:39 ` Jason Gunthorpe
2025-08-15 2:32 ` Shyam Saini
2025-08-18 15:53 ` Jason Gunthorpe
2025-08-26 1:15 ` Shyam Saini
2025-08-26 12:36 ` Jason Gunthorpe
2025-08-26 22:19 ` Shyam Saini
2025-08-06 21:55 ` [PATCH v3 3/3] arm-smmu: select suitable MSI IOVA Shyam Saini
2025-08-06 23:58 ` Jason Gunthorpe
2025-08-07 17:55 ` kernel test robot
2025-08-11 22:34 ` Jacob Pan [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250811153444.000071ab@linux.microsoft.com \
--to=jacob.pan@linux.microsoft.com \
--cc=bboscaccy@linux.microsoft.com \
--cc=code@tyhicks.com \
--cc=devicetree@vger.kernel.org \
--cc=eahariha@linux.microsoft.com \
--cc=eric.auger@redhat.com \
--cc=iommu@lists.linux.dev \
--cc=jgg@ziepe.ca \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=shyamsaini@linux.microsoft.com \
--cc=vijayb@linux.microsoft.com \
--cc=virtualization@lists.linux.dev \
--cc=will@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.