All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Roger Pau Monné" <roger.pau@citrix.com>
To: Jan Beulich <jbeulich@suse.com>
Cc: "xen-devel@lists.xenproject.org" <xen-devel@lists.xenproject.org>,
	Andrew Cooper <andrew.cooper3@citrix.com>,
	Paul Durrant <paul@xen.org>
Subject: Re: [PATCH v2 17/18] AMD/IOMMU: free all-empty page tables
Date: Wed, 15 Dec 2021 16:14:38 +0100	[thread overview]
Message-ID: <YboGXiHaAO1cuUnd@Air-de-Roger> (raw)
In-Reply-To: <ef31b6f4-09f5-cded-a235-5dce3c5399ce@suse.com>

On Fri, Sep 24, 2021 at 11:55:57AM +0200, Jan Beulich wrote:
> When a page table ends up with no present entries left, it can be
> replaced by a non-present entry at the next higher level. The page table
> itself can then be scheduled for freeing.
> 
> Note that while its output isn't used there yet, update_contig_markers()
> right away needs to be called in all places where entries get updated,
> not just the one where entries get cleared.

Couldn't you also coalesce all contiguous page tables into a
super-page entry at the higher level? (not that should be done here,
it's just that I'm not seeing any patch to that effect in the series)

> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> ---
> v2: New.
> 
> --- a/xen/drivers/passthrough/amd/iommu_map.c
> +++ b/xen/drivers/passthrough/amd/iommu_map.c
> @@ -21,6 +21,9 @@
>  
>  #include "iommu.h"
>  
> +#define CONTIG_MASK IOMMU_PTE_CONTIG_MASK
> +#include <asm/contig-marker.h>
> +
>  /* Given pfn and page table level, return pde index */
>  static unsigned int pfn_to_pde_idx(unsigned long pfn, unsigned int level)
>  {
> @@ -33,16 +36,20 @@ static unsigned int pfn_to_pde_idx(unsig
>  
>  static union amd_iommu_pte clear_iommu_pte_present(unsigned long l1_mfn,
>                                                     unsigned long dfn,
> -                                                   unsigned int level)
> +                                                   unsigned int level,
> +                                                   bool *free)
>  {
>      union amd_iommu_pte *table, *pte, old;
> +    unsigned int idx = pfn_to_pde_idx(dfn, level);
>  
>      table = map_domain_page(_mfn(l1_mfn));
> -    pte = &table[pfn_to_pde_idx(dfn, level)];
> +    pte = &table[idx];
>      old = *pte;
>  
>      write_atomic(&pte->raw, 0);
>  
> +    *free = update_contig_markers(&table->raw, idx, level, PTE_kind_null);
> +
>      unmap_domain_page(table);
>  
>      return old;
> @@ -85,7 +92,11 @@ static union amd_iommu_pte set_iommu_pte
>      if ( !old.pr || old.next_level ||
>           old.mfn != next_mfn ||
>           old.iw != iw || old.ir != ir )
> +    {
>          set_iommu_pde_present(pde, next_mfn, 0, iw, ir);
> +        update_contig_markers(&table->raw, pfn_to_pde_idx(dfn, level), level,
> +                              PTE_kind_leaf);
> +    }
>      else
>          old.pr = false; /* signal "no change" to the caller */
>  
> @@ -259,6 +270,9 @@ static int iommu_pde_from_dfn(struct dom
>              smp_wmb();
>              set_iommu_pde_present(pde, next_table_mfn, next_level, true,
>                                    true);
> +            update_contig_markers(&next_table_vaddr->raw,
> +                                  pfn_to_pde_idx(dfn, level),
> +                                  level, PTE_kind_table);
>  
>              *flush_flags |= IOMMU_FLUSHF_modified;
>          }
> @@ -284,6 +298,9 @@ static int iommu_pde_from_dfn(struct dom
>                  next_table_mfn = mfn_x(page_to_mfn(table));
>                  set_iommu_pde_present(pde, next_table_mfn, next_level, true,
>                                        true);
> +                update_contig_markers(&next_table_vaddr->raw,
> +                                      pfn_to_pde_idx(dfn, level),
> +                                      level, PTE_kind_table);

Would be nice if we could pack the update_contig_markers in
set_iommu_pde_present (like you do for clear_iommu_pte_present).

Thanks, Roger.


  reply	other threads:[~2021-12-15 15:15 UTC|newest]

Thread overview: 100+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-24  9:39 [PATCH v2 00/18] IOMMU: superpage support when not sharing pagetables Jan Beulich
2021-09-24  9:41 ` [PATCH v2 01/18] AMD/IOMMU: have callers specify the target level for page table walks Jan Beulich
2021-09-24 10:58   ` Roger Pau Monné
2021-09-24 12:02     ` Jan Beulich
2021-09-24  9:42 ` [PATCH v2 02/18] VT-d: " Jan Beulich
2021-09-24 14:45   ` Roger Pau Monné
2021-09-27  9:04     ` Jan Beulich
2021-09-27  9:13       ` Jan Beulich
2021-11-30 11:56       ` Roger Pau Monné
2021-11-30 14:38         ` Jan Beulich
2021-09-24  9:43 ` [PATCH v2 03/18] IOMMU: have vendor code announce supported page sizes Jan Beulich
2021-11-30 12:25   ` Roger Pau Monné
2021-12-17 14:43   ` Julien Grall
2021-12-21  9:26   ` Rahul Singh
2021-09-24  9:44 ` [PATCH v2 04/18] IOMMU: add order parameter to ->{,un}map_page() hooks Jan Beulich
2021-11-30 13:49   ` Roger Pau Monné
2021-11-30 14:45     ` Jan Beulich
2021-12-17 14:42   ` Julien Grall
2021-09-24  9:45 ` [PATCH v2 05/18] IOMMU: have iommu_{,un}map() split requests into largest possible chunks Jan Beulich
2021-11-30 15:24   ` Roger Pau Monné
2021-12-02 15:59     ` Jan Beulich
2021-09-24  9:46 ` [PATCH v2 06/18] IOMMU/x86: restrict IO-APIC mappings for PV Dom0 Jan Beulich
2021-12-01  9:09   ` Roger Pau Monné
2021-12-01  9:27     ` Jan Beulich
2021-12-01 10:32       ` Roger Pau Monné
2021-12-01 11:45         ` Jan Beulich
2021-12-02 15:12           ` Roger Pau Monné
2021-12-02 15:28             ` Jan Beulich
2021-12-02 19:16               ` Andrew Cooper
2021-12-03  6:41                 ` Jan Beulich
2021-09-24  9:47 ` [PATCH v2 07/18] IOMMU/x86: perform PV Dom0 mappings in batches Jan Beulich
2021-12-02 14:10   ` Roger Pau Monné
2021-12-03 12:38     ` Jan Beulich
2021-12-10  9:36       ` Roger Pau Monné
2021-12-10 11:41         ` Jan Beulich
2021-12-10 12:35           ` Roger Pau Monné
2021-09-24  9:48 ` [PATCH v2 08/18] IOMMU/x86: support freeing of pagetables Jan Beulich
2021-12-02 16:03   ` Roger Pau Monné
2021-12-02 16:10     ` Jan Beulich
2021-12-03  8:30       ` Roger Pau Monné
2021-12-03  9:38         ` Roger Pau Monné
2021-12-03  9:40         ` Jan Beulich
2021-12-10 13:51   ` Roger Pau Monné
2021-12-13  8:38     ` Jan Beulich
2021-09-24  9:48 ` [PATCH v2 09/18] AMD/IOMMU: drop stray TLB flush Jan Beulich
2021-12-02 16:16   ` Roger Pau Monné
2021-09-24  9:51 ` [PATCH v2 10/18] AMD/IOMMU: walk trees upon page fault Jan Beulich
2021-12-03  9:03   ` Roger Pau Monné
2021-12-03  9:49     ` Jan Beulich
2021-12-03  9:55       ` Jan Beulich
2021-12-10 10:23         ` Roger Pau Monné
2021-12-03  9:59     ` Jan Beulich
2021-09-24  9:51 ` [PATCH v2 11/18] AMD/IOMMU: return old PTE from {set,clear}_iommu_pte_present() Jan Beulich
2021-12-10 12:05   ` Roger Pau Monné
2021-12-10 12:59     ` Jan Beulich
2021-12-10 13:53       ` Roger Pau Monné
2021-09-24  9:52 ` [PATCH v2 12/18] AMD/IOMMU: allow use of superpage mappings Jan Beulich
2021-12-10 15:06   ` Roger Pau Monné
2021-12-13  8:49     ` Jan Beulich
2021-12-13  9:45       ` Roger Pau Monné
2021-12-13 10:00         ` Jan Beulich
2021-12-13 10:33           ` Roger Pau Monné
2021-12-13 10:41             ` Jan Beulich
2021-09-24  9:52 ` [PATCH v2 13/18] VT-d: " Jan Beulich
2021-12-13 11:54   ` Roger Pau Monné
2021-12-13 13:39     ` Jan Beulich
2021-09-24  9:53 ` [PATCH v2 14/18] IOMMU: fold flush-all hook into "flush one" Jan Beulich
2021-12-13 15:04   ` Roger Pau Monné
2021-12-14  9:06     ` Jan Beulich
2021-12-14  9:27       ` Roger Pau Monné
2021-12-15 15:28   ` Oleksandr
2021-12-16  8:49     ` Jan Beulich
2021-12-16 10:39       ` Oleksandr
2021-12-16 11:30   ` Rahul Singh
2021-12-21  8:04     ` Jan Beulich
2021-12-17 14:38   ` Julien Grall
2021-09-24  9:54 ` [PATCH v2 15/18] IOMMU/x86: prefill newly allocate page tables Jan Beulich
2021-12-13 15:51   ` Roger Pau Monné
2021-12-14  9:15     ` Jan Beulich
2021-12-14 11:41       ` Roger Pau Monné
2021-12-14 11:48         ` Jan Beulich
2021-12-14 14:50   ` Roger Pau Monné
2021-12-14 15:05     ` Jan Beulich
2021-12-14 15:15       ` Roger Pau Monné
2021-12-14 15:21         ` Jan Beulich
2021-12-14 15:06   ` Roger Pau Monné
2021-12-14 15:10     ` Jan Beulich
2021-12-14 15:17       ` Roger Pau Monné
2021-12-14 15:24         ` Jan Beulich
2021-09-24  9:55 ` [PATCH v2 16/18] x86: introduce helper for recording degree of contiguity in " Jan Beulich
2021-12-15 13:57   ` Roger Pau Monné
2021-12-16 15:47     ` Jan Beulich
2021-12-20 15:25       ` Roger Pau Monné
2021-12-21  8:09         ` Jan Beulich
2022-01-04  8:57           ` Roger Pau Monné
2022-01-04  9:00             ` Jan Beulich
2021-09-24  9:55 ` [PATCH v2 17/18] AMD/IOMMU: free all-empty " Jan Beulich
2021-12-15 15:14   ` Roger Pau Monné [this message]
2021-12-16 15:54     ` Jan Beulich
2021-09-24  9:56 ` [PATCH v2 18/18] VT-d: " Jan Beulich

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=YboGXiHaAO1cuUnd@Air-de-Roger \
    --to=roger.pau@citrix.com \
    --cc=andrew.cooper3@citrix.com \
    --cc=jbeulich@suse.com \
    --cc=paul@xen.org \
    --cc=xen-devel@lists.xenproject.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.