patches.lists.linux.dev archive mirror
 help / color / mirror / Atom feed
From: Jason Gunthorpe <jgg@nvidia.com>
To: Jonathan Corbet <corbet@lwn.net>,
	iommu@lists.linux.dev, Joerg Roedel <joro@8bytes.org>,
	Justin Stitt <justinstitt@google.com>,
	Kevin Tian <kevin.tian@intel.com>,
	linux-doc@vger.kernel.org, linux-kselftest@vger.kernel.org,
	llvm@lists.linux.dev, Bill Wendling <morbo@google.com>,
	Nathan Chancellor <nathan@kernel.org>,
	Nick Desaulniers <nick.desaulniers+lkml@gmail.com>,
	Miguel Ojeda <ojeda@kernel.org>,
	Robin Murphy <robin.murphy@arm.com>,
	Shuah Khan <shuah@kernel.org>,
	Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>,
	Will Deacon <will@kernel.org>
Cc: Alexey Kardashevskiy <aik@amd.com>,
	Alejandro Jimenez <alejandro.j.jimenez@oracle.com>,
	James Gowans <jgowans@amazon.com>,
	Michael Roth <michael.roth@amd.com>,
	Pasha Tatashin <pasha.tatashin@soleen.com>,
	patches@lists.linux.dev
Subject: Re: [PATCH v2 07/15] iommupt: Add map_pages op
Date: Thu, 5 Jun 2025 13:49:26 -0300	[thread overview]
Message-ID: <20250605164926.GA64967@nvidia.com> (raw)
In-Reply-To: <7-v2-5c26bde5c22d+58b-iommu_pt_jgg@nvidia.com>

On Mon, May 05, 2025 at 11:18:37AM -0300, Jason Gunthorpe wrote:
> +/*
> + * Check that the items in a contiguous block are all empty. This will
> + * recursively check any tables in the block to validate they are empty and
> + * accumulate them on the free list. Makes no change on failure. On success
> + * caller must fill the items.
> + */
> +static int clear_contig(const struct pt_state *start_pts,
> +			struct iommu_iotlb_gather *iotlb_gather,
> +			unsigned int step, unsigned int pgsize_lg2)
> +{
> +	struct pt_iommu *iommu_table =
> +		iommu_from_common(start_pts->range->common);
> +	struct pt_range range = *start_pts->range;
> +	struct pt_state pts =
> +		pt_init(&range, start_pts->level, start_pts->table);
> +	struct pt_iommu_collect_args collect = {
> +		.free_list = IOMMU_PAGES_LIST_INIT(collect.free_list),
> +	};
> +	pt_vaddr_t start_va = range.va;
> +	int ret;
> +
> +	pts.index = start_pts->index;
> +	pts.end_index = start_pts->index + step;
> +	for (; _pt_iter_load(&pts); pt_next_entry(&pts)) {
> +		if (pts.type == PT_ENTRY_TABLE) {
> +			ret = pt_walk_descend_all(&pts, __collect_tables,
> +						  &collect);
> +			if (ret)
> +				return ret;
> +
> +			iommu_pages_list_add(&collect.free_list,
> +					     pt_table_ptr(&pts));
> +		} else if (pts.type != PT_ENTRY_EMPTY) {
> +			return -EADDRINUSE;
> +		}
> +	}
> +
> +	if (!iommu_pages_list_empty(&collect.free_list)) {
> +		gather_range(iotlb_gather, iommu_table, start_va,
> +			     range.va - start_va);
> +		iommu_pages_list_splice(&collect.free_list,
> +					&iotlb_gather->freelist);
> +	}

For the PT_FEAT_FLUSH_RANGE_NO_GAPS mode this doesn't work, it could
cause an iommu_iotlb_sync() and free to be done before the PTEs have
been updated. The original idea was that the PTEs would all be set
directly to the new OA PTEs by the caller, which worked OK only for
PT_FEAT_FLUSH_RANGE.

It looks like fixing it by clearing the table levels as soon as we
know they are empty is the best thing:

+                       /*
+                        * The table item must be cleared before we can update
+                        * the gather
+                        */
+                       pt_clear_entry(&pts, ilog2(1));
+
                        iommu_pages_list_add(&collect.free_list,
                                             pt_table_ptr(&pts));
+                       gather_range_pages(
+                               iotlb_gather, iommu_table, range.va,
+                               log2_to_int(pt_table_item_lg2sz(&pts)),
+                               &collect.free_list);

Jason

  parent reply	other threads:[~2025-06-05 16:49 UTC|newest]

Thread overview: 38+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-05-05 14:18 [PATCH v2 00/15] Consolidate iommu page table implementations (AMD) Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 01/15] genpt: Generic Page Table base API Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 02/15] genpt: Add Documentation/ files Jason Gunthorpe
2025-05-07  2:37   ` Bagas Sanjaya
2025-05-13 18:53     ` Jason Gunthorpe
2025-05-15  6:21       ` Bagas Sanjaya
2025-05-05 14:18 ` [PATCH v2 03/15] iommupt: Add the basic structure of the iommu implementation Jason Gunthorpe
2025-05-14  9:23   ` Ankit Soni
2025-05-14 15:54     ` Jason Gunthorpe
2025-05-14 20:08       ` Alejandro Jimenez
2025-05-15 19:32         ` Jason Gunthorpe
2025-05-16  5:02         ` Ankit Soni
2025-05-16 20:39           ` Alejandro Jimenez
2025-05-05 14:18 ` [PATCH v2 04/15] iommupt: Add the AMD IOMMU v1 page table format Jason Gunthorpe
2025-05-05 18:48   ` ALOK TIWARI
2025-05-16  8:30   ` Yi Liu
2025-05-16 11:57     ` Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 05/15] iommupt: Add iova_to_phys op Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 06/15] iommupt: Add unmap_pages op Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 07/15] iommupt: Add map_pages op Jason Gunthorpe
2025-05-13  5:15   ` Ankit Soni
2025-05-13 12:00     ` Jason Gunthorpe
2025-06-05 16:49   ` Jason Gunthorpe [this message]
2025-05-05 14:18 ` [PATCH v2 08/15] iommupt: Add read_and_clear_dirty op Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 09/15] iommupt: Add a kunit test for Generic Page Table Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 10/15] iommupt: Add a mock pagetable format for iommufd selftest to use Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 11/15] iommufd: Change the selftest to use iommupt instead of xarray Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 12/15] iommupt: Add the x86 64 bit page table format Jason Gunthorpe
2025-06-05 21:03   ` Jacob Pan
2025-06-06 11:43     ` Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 13/15] iommu/amd: Use the generic iommu page table Jason Gunthorpe
2025-05-05 19:09   ` ALOK TIWARI
2025-05-05 14:18 ` [PATCH v2 14/15] iommu/amd: Remove AMD io_pgtable support Jason Gunthorpe
2025-05-05 14:18 ` [PATCH v2 15/15] iommupt: Add a kunit test for the IOMMU implementation Jason Gunthorpe
2025-05-05 17:44   ` Nathan Chancellor
2025-05-05 17:47     ` Jason Gunthorpe
2025-05-05 18:00       ` Nathan Chancellor
2025-05-13  1:08 ` [PATCH v2 00/15] Consolidate iommu page table implementations (AMD) Alejandro Jimenez

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=20250605164926.GA64967@nvidia.com \
    --to=jgg@nvidia.com \
    --cc=aik@amd.com \
    --cc=alejandro.j.jimenez@oracle.com \
    --cc=corbet@lwn.net \
    --cc=iommu@lists.linux.dev \
    --cc=jgowans@amazon.com \
    --cc=joro@8bytes.org \
    --cc=justinstitt@google.com \
    --cc=kevin.tian@intel.com \
    --cc=linux-doc@vger.kernel.org \
    --cc=linux-kselftest@vger.kernel.org \
    --cc=llvm@lists.linux.dev \
    --cc=michael.roth@amd.com \
    --cc=morbo@google.com \
    --cc=nathan@kernel.org \
    --cc=nick.desaulniers+lkml@gmail.com \
    --cc=ojeda@kernel.org \
    --cc=pasha.tatashin@soleen.com \
    --cc=patches@lists.linux.dev \
    --cc=robin.murphy@arm.com \
    --cc=shuah@kernel.org \
    --cc=suravee.suthikulpanit@amd.com \
    --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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).