[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v5 06/10] AMD/IOMMU: don't blindly allocate interrupt remapping tables


  • To: Jan Beulich <jbeulich@xxxxxxxx>, "xen-devel@xxxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • From: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
  • Date: Tue, 17 Sep 2019 14:10:41 +0100
  • Authentication-results: esa4.hc3370-68.iphmx.com; dkim=none (message not signed) header.i=none; spf=None smtp.pra=andrew.cooper3@xxxxxxxxxx; spf=Pass smtp.mailfrom=Andrew.Cooper3@xxxxxxxxxx; spf=None smtp.helo=postmaster@xxxxxxxxxxxxxxx
  • Autocrypt: addr=andrew.cooper3@xxxxxxxxxx; prefer-encrypt=mutual; keydata= mQINBFLhNn8BEADVhE+Hb8i0GV6mihnnr/uiQQdPF8kUoFzCOPXkf7jQ5sLYeJa0cQi6Penp VtiFYznTairnVsN5J+ujSTIb+OlMSJUWV4opS7WVNnxHbFTPYZVQ3erv7NKc2iVizCRZ2Kxn srM1oPXWRic8BIAdYOKOloF2300SL/bIpeD+x7h3w9B/qez7nOin5NzkxgFoaUeIal12pXSR Q354FKFoy6Vh96gc4VRqte3jw8mPuJQpfws+Pb+swvSf/i1q1+1I4jsRQQh2m6OTADHIqg2E ofTYAEh7R5HfPx0EXoEDMdRjOeKn8+vvkAwhviWXTHlG3R1QkbE5M/oywnZ83udJmi+lxjJ5 YhQ5IzomvJ16H0Bq+TLyVLO/VRksp1VR9HxCzItLNCS8PdpYYz5TC204ViycobYU65WMpzWe LFAGn8jSS25XIpqv0Y9k87dLbctKKA14Ifw2kq5OIVu2FuX+3i446JOa2vpCI9GcjCzi3oHV e00bzYiHMIl0FICrNJU0Kjho8pdo0m2uxkn6SYEpogAy9pnatUlO+erL4LqFUO7GXSdBRbw5 gNt25XTLdSFuZtMxkY3tq8MFss5QnjhehCVPEpE6y9ZjI4XB8ad1G4oBHVGK5LMsvg22PfMJ ISWFSHoF/B5+lHkCKWkFxZ0gZn33ju5n6/FOdEx4B8cMJt+cWwARAQABtClBbmRyZXcgQ29v cGVyIDxhbmRyZXcuY29vcGVyM0BjaXRyaXguY29tPokCOgQTAQgAJAIbAwULCQgHAwUVCgkI CwUWAgMBAAIeAQIXgAUCWKD95wIZAQAKCRBlw/kGpdefoHbdD/9AIoR3k6fKl+RFiFpyAhvO 59ttDFI7nIAnlYngev2XUR3acFElJATHSDO0ju+hqWqAb8kVijXLops0gOfqt3VPZq9cuHlh IMDquatGLzAadfFx2eQYIYT+FYuMoPZy/aTUazmJIDVxP7L383grjIkn+7tAv+qeDfE+txL4 SAm1UHNvmdfgL2/lcmL3xRh7sub3nJilM93RWX1Pe5LBSDXO45uzCGEdst6uSlzYR/MEr+5Z JQQ32JV64zwvf/aKaagSQSQMYNX9JFgfZ3TKWC1KJQbX5ssoX/5hNLqxMcZV3TN7kU8I3kjK mPec9+1nECOjjJSO/h4P0sBZyIUGfguwzhEeGf4sMCuSEM4xjCnwiBwftR17sr0spYcOpqET ZGcAmyYcNjy6CYadNCnfR40vhhWuCfNCBzWnUW0lFoo12wb0YnzoOLjvfD6OL3JjIUJNOmJy RCsJ5IA/Iz33RhSVRmROu+TztwuThClw63g7+hoyewv7BemKyuU6FTVhjjW+XUWmS/FzknSi dAG+insr0746cTPpSkGl3KAXeWDGJzve7/SBBfyznWCMGaf8E2P1oOdIZRxHgWj0zNr1+ooF /PzgLPiCI4OMUttTlEKChgbUTQ+5o0P080JojqfXwbPAyumbaYcQNiH1/xYbJdOFSiBv9rpt TQTBLzDKXok86LkCDQRS4TZ/ARAAkgqudHsp+hd82UVkvgnlqZjzz2vyrYfz7bkPtXaGb9H4 Rfo7mQsEQavEBdWWjbga6eMnDqtu+FC+qeTGYebToxEyp2lKDSoAsvt8w82tIlP/EbmRbDVn 7bhjBlfRcFjVYw8uVDPptT0TV47vpoCVkTwcyb6OltJrvg/QzV9f07DJswuda1JH3/qvYu0p vjPnYvCq4NsqY2XSdAJ02HrdYPFtNyPEntu1n1KK+gJrstjtw7KsZ4ygXYrsm/oCBiVW/OgU g/XIlGErkrxe4vQvJyVwg6YH653YTX5hLLUEL1NS4TCo47RP+wi6y+TnuAL36UtK/uFyEuPy wwrDVcC4cIFhYSfsO0BumEI65yu7a8aHbGfq2lW251UcoU48Z27ZUUZd2Dr6O/n8poQHbaTd 6bJJSjzGGHZVbRP9UQ3lkmkmc0+XCHmj5WhwNNYjgbbmML7y0fsJT5RgvefAIFfHBg7fTY/i kBEimoUsTEQz+N4hbKwo1hULfVxDJStE4sbPhjbsPCrlXf6W9CxSyQ0qmZ2bXsLQYRj2xqd1 bpA+1o1j2N4/au1R/uSiUFjewJdT/LX1EklKDcQwpk06Af/N7VZtSfEJeRV04unbsKVXWZAk uAJyDDKN99ziC0Wz5kcPyVD1HNf8bgaqGDzrv3TfYjwqayRFcMf7xJaL9xXedMcAEQEAAYkC HwQYAQgACQUCUuE2fwIbDAAKCRBlw/kGpdefoG4XEACD1Qf/er8EA7g23HMxYWd3FXHThrVQ HgiGdk5Yh632vjOm9L4sd/GCEACVQKjsu98e8o3ysitFlznEns5EAAXEbITrgKWXDDUWGYxd pnjj2u+GkVdsOAGk0kxczX6s+VRBhpbBI2PWnOsRJgU2n10PZ3mZD4Xu9kU2IXYmuW+e5KCA vTArRUdCrAtIa1k01sPipPPw6dfxx2e5asy21YOytzxuWFfJTGnVxZZSCyLUO83sh6OZhJkk b9rxL9wPmpN/t2IPaEKoAc0FTQZS36wAMOXkBh24PQ9gaLJvfPKpNzGD8XWR5HHF0NLIJhgg 4ZlEXQ2fVp3XrtocHqhu4UZR4koCijgB8sB7Tb0GCpwK+C4UePdFLfhKyRdSXuvY3AHJd4CP 4JzW0Bzq/WXY3XMOzUTYApGQpnUpdOmuQSfpV9MQO+/jo7r6yPbxT7CwRS5dcQPzUiuHLK9i nvjREdh84qycnx0/6dDroYhp0DFv4udxuAvt1h4wGwTPRQZerSm4xaYegEFusyhbZrI0U9tJ B8WrhBLXDiYlyJT6zOV2yZFuW47VrLsjYnHwn27hmxTC/7tvG3euCklmkn9Sl9IAKFu29RSo d5bD8kMSCYsTqtTfT6W4A3qHGvIDta3ptLYpIAOD2sY3GYq2nf3Bbzx81wZK14JdDDHUX2Rs 6+ahAA==
  • Cc: Suravee Suthikulpanit <suravee.suthikulpanit@xxxxxxx>
  • Delivery-date: Tue, 17 Sep 2019 13:10:53 +0000
  • Ironport-sdr: +IT9N7slrpFbH5shk5tYMir44E/B2mrc9HS+i6Mf5a1hCrpH1rYtUJql2Y8wQoh6ValD9InDsi mkCaLrlxjQMqoowYOwBN4/T6ZdDgQ3+lNXPijSQLDNHT8yJWtfSvyEBKIVZQ1j1Wp5tNXF6v4O XdCOZOfy7eGCLCe7xv5EsDhdqeFW0CCbGiMc+AgS/8ze/iRamxdRXjIWuAWKQ5Z4ygHA/R6kS/ esLh6UgCJyi5TovJDx98QL8eA78lAcjpTmNfDz0y+SEgbJh79ACZ6XuHxC2HyDMZwQv+qvvzgQ qks=
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>
  • Openpgp: preference=signencrypt

On 06/08/2019 14:09, Jan Beulich wrote:
> ACPI tables are free to list far more device coordinates than there are
> actual devices. By delaying the table allocations for most cases, and
> doing them only when an actual device is known to be present at a given
> position, overall memory used for the tables goes down from over 500k
> pages to just over 1k ones.

This is slightly awkward grammar.  While I don't think it is strictly
speaking incorrect, it would be more normal to phrase as "just over 1k
pages".

>
> Tables continue to get allocated right away for special entries
> (IO-APIC, HPET) as well as for alias IDs. While in the former case
> that's simply because there may not be any device at a given position,
> in the latter case this is to avoid having to introduce ref-counting of
> table usage.
>
> The change involves invoking
> iterate_ivrs_mappings(amd_iommu_setup_device_table) a second time,
> because the function now wants to be able to find PCI devices, which
> isn't possible yet when IOMMU setup happens very early during x2APIC
> mode setup. In this context amd_iommu_init_interrupt() gets renamed as
> well.
>
> The logic adjusting a DTE's interrupt remapping attributes also gets
> changed, such that the lack of an IRT would result in target aborted
> rather than not remapped interrupts (should any occur).
>
> Note that for now phantom functions get separate IRTs allocated, as was
> the case before.
>
> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> ---
> v5: New.
> ---
> TBD: This retains prior (but suspicious) behavior of not calling
>      amd_iommu_set_intremap_table() for "invalid" IVRS mapping entries.
>      Since DTE.IV=0 means un-remapped interrupts, I wonder if this needs
>      changing.

How would an invalid entry get DTE.V set in the first place?

Whatever the old behaviour may have been, we shouldn't have code which
comes with a chance of having non-remapped interrupts by accident.

>
> @@ -1266,13 +1270,46 @@ static int __init amd_iommu_setup_device
>          if ( ivrs_mappings[bdf].valid )
>          {
>              void *dte;
> +            const struct pci_dev *pdev = NULL;
>  
>              /* add device table entry */
>              dte = device_table.buffer + (bdf *
> IOMMU_DEV_TABLE_ENTRY_SIZE);
>              iommu_dte_add_device_entry(dte, &ivrs_mappings[bdf]);
>  
> +            if ( iommu_intremap &&
> +                 ivrs_mappings[bdf].dte_requestor_id == bdf &&
> +                 !ivrs_mappings[bdf].intremap_table )
> +            {
> +                if ( !pci_init )
> +                    continue;
> +                pcidevs_lock();
> +                pdev = pci_get_pdev(seg, PCI_BUS(bdf), PCI_DEVFN2(bdf));
> +                pcidevs_unlock();
> +            }
> +
> +            if ( pdev )
> +            {
> +                unsigned int req_id = bdf;
> +
> +                do {
> +                    ivrs_mappings[req_id].intremap_table =
> +                        amd_iommu_alloc_intremap_table(
> +                            ivrs_mappings[bdf].iommu,
> +                            &ivrs_mappings[req_id].intremap_inuse);
> +                    if ( !ivrs_mappings[req_id].intremap_table )
> +                        return -ENOMEM;
> +
> +                    if ( !pdev->phantom_stride )
> +                        break;
> +                    req_id += pdev->phantom_stride;
> +                } while ( PCI_SLOT(req_id) == pdev->sbdf.dev );
> +            }
> +
>              amd_iommu_set_intremap_table(
> -                dte, virt_to_maddr(ivrs_mappings[bdf].intremap_table),
> +                dte,
> +                ivrs_mappings[bdf].intremap_table
> +                ? virt_to_maddr(ivrs_mappings[bdf].intremap_table)
> +                : 0,

Under what circumstances can ivrs_mappings[bdf].intremap_table be NULL,
given the ENOMEM above?

This case isn't very clear cut given the !pdev possibility, but...

> --- a/xen/drivers/passthrough/amd/pci_amd_iommu.c
> +++ b/xen/drivers/passthrough/amd/pci_amd_iommu.c
> @@ -458,6 +459,32 @@ static int amd_iommu_add_device(u8 devfn
>          return -ENODEV;
>      }
>  
> +    ivrs_mappings = get_ivrs_mappings(pdev->seg);
> +    bdf = PCI_BDF2(pdev->bus, devfn);
> +    if ( !ivrs_mappings ||
> +         !ivrs_mappings[ivrs_mappings[bdf].dte_requestor_id].valid )
> +        return -EPERM;
> +
> +    if ( iommu_intremap &&
> +         ivrs_mappings[bdf].dte_requestor_id == bdf &&
> +         !ivrs_mappings[bdf].intremap_table )
> +    {
> +        ivrs_mappings[bdf].intremap_table =
> +            amd_iommu_alloc_intremap_table(
> +                iommu, &ivrs_mappings[bdf].intremap_inuse);
> +        if ( !ivrs_mappings[bdf].intremap_table )
> +            return -ENOMEM;
> +
> +        amd_iommu_set_intremap_table(
> +            iommu->dev_table.buffer + (bdf *
> IOMMU_DEV_TABLE_ENTRY_SIZE),
> +            ivrs_mappings[bdf].intremap_table
> +            ? virt_to_maddr(ivrs_mappings[bdf].intremap_table)
> +            : 0,

... this case definitely cannot occur.

~Andrew

> +            iommu_intremap);
> +
> +        amd_iommu_flush_device(iommu, bdf);
> +    }
> +
>      amd_iommu_setup_domain_device(pdev->domain, iommu, devfn, pdev);
>      return 0;
>  }
>


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.