qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: David Gibson <david@gibson.dropbear.id.au>
To: Alexey Kardashevskiy <aik@ozlabs.ru>
Cc: Alex Williamson <alex.williamson@redhat.com>,
	qemu-ppc@nongnu.org, qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [Qemu-ppc] [PATCH qemu v13 04/16] spapr_iommu: Introduce "enabled" state for TCE table
Date: Thu, 3 Mar 2016 14:00:02 +1100	[thread overview]
Message-ID: <20160303030002.GE1620@voom.redhat.com> (raw)
In-Reply-To: <1456823441-46757-5-git-send-email-aik@ozlabs.ru>

[-- Attachment #1: Type: text/plain, Size: 11633 bytes --]

On Tue, Mar 01, 2016 at 08:10:29PM +1100, Alexey Kardashevskiy wrote:
> Currently TCE tables are created once at start and their sizes never
> change. We are going to change that by introducing a Dynamic DMA windows
> support where DMA configuration may change during the guest execution.
> 
> This changes spapr_tce_new_table() to create an empty zero-size IOMMU
> memory region (IOMMU MR). Only LIOBN is assigned by the time of creation.
> It still will be called once at the owner object (VIO or PHB) creation.
> 
> This introduces an "enabled" state for TCE table objects with two
> helper functions - spapr_tce_table_enable()/spapr_tce_table_disable().
> - spapr_tce_table_enable() receives TCE table parameters, allocates
> a guest view of the TCE table (in the user space or KVM) and
> sets the correct size on the IOMMU MR.
> - spapr_tce_table_disable() disposes the table and resets the IOMMU MR
> size.
> 
> This changes the PHB reset handler to do the default DMA initialization
> instead of spapr_phb_realize(). This does not make differenct now but
> later with more than just one DMA window, we will have to remove them all
> and create the default one on a system reset.
> 
> No visible change in behaviour is expected except the actual table
> will be reallocated every reset. We might optimize this later.
> 
> The other way to implement this would be dynamically create/remove
> the TCE table QOM objects but this would make migration impossible
> as the migration code expects all QOM objects to exist at the receiver
> so we have to have TCE table objects created when migration begins.
> 
> spapr_tce_table_do_enable() is separated from from spapr_tce_table_enable()
> as later it will be called at the sPAPRTCETable post-migration stage when
> it already has all the properties set after the migration.
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

Although there's one nit that could be improved:


> ---
>  hw/ppc/spapr_iommu.c   | 80 +++++++++++++++++++++++++++++++++++---------------
>  hw/ppc/spapr_pci.c     | 21 +++++++++----
>  hw/ppc/spapr_vio.c     |  9 +++---
>  include/hw/ppc/spapr.h | 10 +++----
>  4 files changed, 80 insertions(+), 40 deletions(-)
> 
> diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
> index 8132f64..e66e128 100644
> --- a/hw/ppc/spapr_iommu.c
> +++ b/hw/ppc/spapr_iommu.c
> @@ -174,15 +174,8 @@ static int spapr_tce_table_realize(DeviceState *dev)
>      sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
>  
>      tcet->fd = -1;
> -    tcet->table = spapr_tce_alloc_table(tcet->liobn,
> -                                        tcet->page_shift,
> -                                        tcet->nb_table,
> -                                        &tcet->fd,
> -                                        tcet->need_vfio);
> -
>      memory_region_init_iommu(&tcet->iommu, OBJECT(dev), &spapr_iommu_ops,
> -                             "iommu-spapr",
> -                             (uint64_t)tcet->nb_table << tcet->page_shift);
> +                             "iommu-spapr", 0);
>  
>      QLIST_INSERT_HEAD(&spapr_tce_tables, tcet, list);
>  
> @@ -224,14 +217,10 @@ void spapr_tce_set_need_vfio(sPAPRTCETable *tcet, bool need_vfio)
>      tcet->table = newtable;
>  }
>  
> -sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn,
> -                                   uint64_t bus_offset,
> -                                   uint32_t page_shift,
> -                                   uint32_t nb_table,
> -                                   bool need_vfio)
> +sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn)
>  {
>      sPAPRTCETable *tcet;
> -    char tmp[64];
> +    char tmp[32];
>  
>      if (spapr_tce_find_by_liobn(liobn)) {
>          fprintf(stderr, "Attempted to create TCE table with duplicate"
> @@ -239,16 +228,8 @@ sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn,
>          return NULL;
>      }
>  
> -    if (!nb_table) {
> -        return NULL;
> -    }
> -
>      tcet = SPAPR_TCE_TABLE(object_new(TYPE_SPAPR_TCE_TABLE));
>      tcet->liobn = liobn;
> -    tcet->bus_offset = bus_offset;
> -    tcet->page_shift = page_shift;
> -    tcet->nb_table = nb_table;
> -    tcet->need_vfio = need_vfio;
>  
>      snprintf(tmp, sizeof(tmp), "tce-table-%x", liobn);
>      object_property_add_child(OBJECT(owner), tmp, OBJECT(tcet), NULL);
> @@ -258,14 +239,65 @@ sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn,
>      return tcet;
>  }
>  
> +static void spapr_tce_table_do_enable(sPAPRTCETable *tcet)
> +{
> +    if (!tcet->nb_table) {
> +        return;
> +    }
> +
> +    tcet->table = spapr_tce_alloc_table(tcet->liobn,
> +                                        tcet->page_shift,
> +                                        tcet->nb_table,
> +                                        &tcet->fd,
> +                                        tcet->need_vfio);
> +
> +    memory_region_set_size(&tcet->iommu,
> +                           (uint64_t)tcet->nb_table << tcet->page_shift);
> +
> +    tcet->enabled = true;
> +}
> +
> +void spapr_tce_table_enable(sPAPRTCETable *tcet,
> +                            uint32_t page_shift, uint64_t bus_offset,
> +                            uint32_t nb_table, bool need_vfio)
> +{
> +    if (tcet->enabled) {
> +        return;

If the given parameters are different from the current ones, treating
this as a no-op is rather misleading.  I gather that to resize the
window you're expected to disable, then re-enable.  In which case I
think it would be safer to actually throw some kind of error on a
double enable.


> +    }
> +
> +    tcet->bus_offset = bus_offset;
> +    tcet->page_shift = page_shift;
> +    tcet->nb_table = nb_table;
> +    tcet->need_vfio = need_vfio;
> +
> +    spapr_tce_table_do_enable(tcet);
> +}
> +
> +static void spapr_tce_table_disable(sPAPRTCETable *tcet)
> +{
> +    if (!tcet->enabled) {
> +        return;
> +    }
> +
> +    memory_region_set_size(&tcet->iommu, 0);
> +
> +    spapr_tce_free_table(tcet->table, tcet->fd, tcet->nb_table);
> +    tcet->fd = -1;
> +    tcet->table = NULL;
> +    tcet->enabled = false;
> +    tcet->bus_offset = 0;
> +    tcet->page_shift = 0;
> +    tcet->nb_table = 0;
> +    tcet->need_vfio = false;
> +}
> +
>  static void spapr_tce_table_unrealize(DeviceState *dev, Error **errp)
>  {
>      sPAPRTCETable *tcet = SPAPR_TCE_TABLE(dev);
>  
>      QLIST_REMOVE(tcet, list);
>  
> -    spapr_tce_free_table(tcet->table, tcet->fd, tcet->nb_table);
> -    tcet->fd = -1;
> +    spapr_tce_table_disable(tcet);
>  }
>  
>  MemoryRegion *spapr_tce_get_iommu(sPAPRTCETable *tcet)
> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
> index 248f20a..c34a906 100644
> --- a/hw/ppc/spapr_pci.c
> +++ b/hw/ppc/spapr_pci.c
> @@ -815,12 +815,13 @@ static int spapr_phb_dma_window_enable(sPAPRPHBState *sphb,
>          return -1;
>      }
>  
> -    tcet = spapr_tce_new_table(DEVICE(sphb), liobn, window_addr,
> -                               page_shift, nb_table, false);
> +    tcet = spapr_tce_find_by_liobn(liobn);
>      if (!tcet) {
>          return -1;
>      }
>  
> +    spapr_tce_table_enable(tcet, page_shift, window_addr, nb_table, false);
> +
>      memory_region_add_subregion(&sphb->iommu_root, tcet->bus_offset,
>                                  spapr_tce_get_iommu(tcet));
>      return 0;
> @@ -1251,6 +1252,7 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>      int i;
>      PCIBus *bus;
>      uint64_t msi_window_size = 4096;
> +    sPAPRTCETable *tcet;
>  
>      if (sphb->index != (uint32_t)-1) {
>          hwaddr windows_base;
> @@ -1402,11 +1404,18 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>          }
>      }
>  
> +    /* DMA setup */
> +    tcet = spapr_tce_new_table(DEVICE(sphb), sphb->dma_liobn);
> +    if (!tcet) {
> +        error_report("No default TCE table for %s", sphb->dtbusname);
> +        return;
> +    }
> +
>      /* Register default 32bit DMA window */
> -    if (spapr_phb_dma_window_enable(sphb, sphb->dma_liobn, SPAPR_TCE_PAGE_SHIFT,
> -                                    sphb->dma_win_addr, sphb->dma_win_size)) {
> -        error_setg(errp, "Unable to create TCE table for %s", sphb->dtbusname);
> -    }
> +    spapr_phb_dma_window_enable(sphb, sphb->dma_liobn,
> +                                SPAPR_TCE_PAGE_SHIFT,
> +                                sphb->dma_win_addr,
> +                                sphb->dma_win_size);
>  
>      sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
>  }
> diff --git a/hw/ppc/spapr_vio.c b/hw/ppc/spapr_vio.c
> index 0f61a55..a745884 100644
> --- a/hw/ppc/spapr_vio.c
> +++ b/hw/ppc/spapr_vio.c
> @@ -481,11 +481,10 @@ static void spapr_vio_busdev_realize(DeviceState *qdev, Error **errp)
>          memory_region_add_subregion_overlap(&dev->mrroot, 0, &dev->mrbypass, 1);
>          address_space_init(&dev->as, &dev->mrroot, qdev->id);
>  
> -        dev->tcet = spapr_tce_new_table(qdev, liobn,
> -                                        0,
> -                                        SPAPR_TCE_PAGE_SHIFT,
> -                                        pc->rtce_window_size >>
> -                                        SPAPR_TCE_PAGE_SHIFT, false);
> +        dev->tcet = spapr_tce_new_table(qdev, liobn);
> +        spapr_tce_table_enable(dev->tcet, SPAPR_TCE_PAGE_SHIFT, 0,
> +                               pc->rtce_window_size >> SPAPR_TCE_PAGE_SHIFT,
> +                               false);
>          dev->tcet->vdev = dev;
>          memory_region_add_subregion_overlap(&dev->mrroot, 0,
>                                              spapr_tce_get_iommu(dev->tcet), 2);
> diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
> index 098d85d..3e6bb84 100644
> --- a/include/hw/ppc/spapr.h
> +++ b/include/hw/ppc/spapr.h
> @@ -539,6 +539,7 @@ typedef struct sPAPRTCETable sPAPRTCETable;
>  
>  struct sPAPRTCETable {
>      DeviceState parent;
> +    bool enabled;
>      uint32_t liobn;
>      uint32_t nb_table;
>      uint64_t bus_offset;
> @@ -566,11 +567,10 @@ void spapr_events_fdt_skel(void *fdt, uint32_t epow_irq);
>  int spapr_h_cas_compose_response(sPAPRMachineState *sm,
>                                   target_ulong addr, target_ulong size,
>                                   bool cpu_update, bool memory_update);
> -sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn,
> -                                   uint64_t bus_offset,
> -                                   uint32_t page_shift,
> -                                   uint32_t nb_table,
> -                                   bool need_vfio);
> +sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn);
> +void spapr_tce_table_enable(sPAPRTCETable *tcet,
> +                            uint32_t page_shift, uint64_t bus_offset,
> +                            uint32_t nb_table, bool vfio_accel);
>  void spapr_tce_set_need_vfio(sPAPRTCETable *tcet, bool need_vfio);
>  
>  MemoryRegion *spapr_tce_get_iommu(sPAPRTCETable *tcet);

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

  reply	other threads:[~2016-03-03  3:05 UTC|newest]

Thread overview: 50+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-03-01  9:10 [Qemu-devel] [PATCH qemu v13 00/16] spapr: vfio: Enable Dynamic DMA windows (DDW) Alexey Kardashevskiy
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 01/16] memory: Fix IOMMU replay base address Alexey Kardashevskiy
2016-03-03  1:34   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 02/16] spapr_pci: Move DMA window enablement to a helper Alexey Kardashevskiy
2016-03-03  1:40   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-10  5:47     ` Alexey Kardashevskiy
2016-03-15  5:30       ` David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 03/16] spapr_iommu: Move table allocation to helpers Alexey Kardashevskiy
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 04/16] spapr_iommu: Introduce "enabled" state for TCE table Alexey Kardashevskiy
2016-03-03  3:00   ` David Gibson [this message]
2016-03-10  7:39     ` [Qemu-devel] [Qemu-ppc] " Alexey Kardashevskiy
2016-03-15  5:32       ` David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 05/16] spapr_iommu: Add root memory region Alexey Kardashevskiy
2016-03-04  4:08   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 06/16] spapr_pci: Reset DMA config on PHB reset Alexey Kardashevskiy
2016-03-03  3:02   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 07/16] vfio, memory: Notify IOMMU about starting/stopping being used by VFIO Alexey Kardashevskiy
2016-03-03  5:28   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-03  6:01     ` Alexey Kardashevskiy
2016-03-04  4:01       ` David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 08/16] memory: Add reporting of supported page sizes Alexey Kardashevskiy
2016-03-03  5:33   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 09/16] vfio: Generalize IOMMU memory listener Alexey Kardashevskiy
2016-03-03  5:36   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-03  6:07     ` Alexey Kardashevskiy
2016-03-04  3:44       ` David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 10/16] vfio: Use different page size for different IOMMU types Alexey Kardashevskiy
2016-03-03  6:08   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 11/16] vfio: spapr: Add SPAPR IOMMU v2 support (DMA memory preregistering) Alexey Kardashevskiy
2016-03-03  6:30   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-15  2:53     ` Alexey Kardashevskiy
2016-03-15  5:42       ` David Gibson
2016-03-17  5:04         ` Alexey Kardashevskiy
2016-03-17  6:10           ` David Gibson
2016-03-17  9:23             ` Alexey Kardashevskiy
2016-03-21  4:53               ` David Gibson
2016-03-21  6:08                 ` Alexey Kardashevskiy
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 12/16] vmstate: Define VARRAY with VMS_ALLOC Alexey Kardashevskiy
2016-03-03  6:31   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 13/16] spapr_iommu: Remove need_vfio flag from sPAPRTCETable Alexey Kardashevskiy
2016-03-03  6:38   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 14/16] spapr_pci: Add and export DMA resetting helper Alexey Kardashevskiy
2016-03-03  6:39   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 15/16] vfio: Move iova_pgsizes from container to guest IOMMU Alexey Kardashevskiy
2016-03-03 11:22   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-04  0:02     ` Alexey Kardashevskiy
2016-03-01  9:10 ` [Qemu-devel] [PATCH qemu v13 16/16] spapr_pci/spapr_pci_vfio: Support Dynamic DMA Windows (DDW) Alexey Kardashevskiy
2016-03-04  4:51   ` [Qemu-devel] [Qemu-ppc] " David Gibson
2016-03-11  9:03     ` Alexey Kardashevskiy
2016-03-15  5:53       ` David Gibson

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=20160303030002.GE1620@voom.redhat.com \
    --to=david@gibson.dropbear.id.au \
    --cc=aik@ozlabs.ru \
    --cc=alex.williamson@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=qemu-ppc@nongnu.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).