[Qemu-devel] [RFC PATCH] spapr_pci: make index property mandatory

Greg Kurz posted 1 patch 6 years, 7 months ago
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/150512970330.16081.7553039475680302776.stgit@bahia
Test checkpatch failed
Test docker passed
Test s390x passed
There is a newer version of this series
hw/ppc/spapr_pci.c |   52 ++++++++++------------------------------------------
1 file changed, 10 insertions(+), 42 deletions(-)
[Qemu-devel] [RFC PATCH] spapr_pci: make index property mandatory
Posted by Greg Kurz 6 years, 7 months ago
Creating several PHBs without index property confuses the DRC code
and causes issues:
- only the first index-less PHB is functional, the other ones will
  silently ignore hotplugging of PCI devices
- QEMU will even terminate if these PHBs have cold-plugged devices

qemu-system-ppc64: -device virtio-net,bus=pci2.0: an attached device
 is still awaiting release

This happens because DR connectors for child PCI devices are created
with a DRC index that is derived from the PHB's index property. If the
PHBs are created without index, then the same value of -1 is used to
compute the DRC indexes for both PHBs, hence causing the collision.

Also, the index property is used to compute the placement of the PHB's
memory regions. It is limited to 31 or 255, depending on the machine
type version. This fits well with the requirements of DRC indexes, which
need the PHB index to be a 16-bit value.

This patch hence makes the index property mandatory. As a consequence,
the PHB's memory regions and BUID are now always configured according
to the index, and it is no longer possible to set them from the command
line. This is a reasonable trade-off, as it is very unlikely that people
create PHBs without index (at least libvirt doesn't do it FWIW).

We have to introduce a PHB instance init function to initialize the
64-bit window address to -1 because pseries-2.7 and older machines
don't set it.

Signed-off-by: Greg Kurz <groug@kaod.org>
---

Hi,

This is a proposal to address the issue uncovered during the review of the
PHB hotplug patches:

https://lists.nongnu.org/archive/html/qemu-ppc/2017-07/msg00557.html

I'd like to address this properly before resuming work on PHB hotplug
itself. Please comment.

Cheers,

--
Greg
---
 hw/ppc/spapr_pci.c |   52 ++++++++++------------------------------------------
 1 file changed, 10 insertions(+), 42 deletions(-)

diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index cf54160526fa..9a338b7f197b 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -1523,16 +1523,6 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
         sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
         Error *local_err = NULL;
 
-        if ((sphb->buid != (uint64_t)-1) || (sphb->dma_liobn[0] != (uint32_t)-1)
-            || (sphb->dma_liobn[1] != (uint32_t)-1 && windows_supported == 2)
-            || (sphb->mem_win_addr != (hwaddr)-1)
-            || (sphb->mem64_win_addr != (hwaddr)-1)
-            || (sphb->io_win_addr != (hwaddr)-1)) {
-            error_setg(errp, "Either \"index\" or other parameters must"
-                       " be specified for PAPR PHB, not both");
-            return;
-        }
-
         smc->phb_placement(spapr, sphb->index,
                            &sphb->buid, &sphb->io_win_addr,
                            &sphb->mem_win_addr, &sphb->mem64_win_addr,
@@ -1541,36 +1531,12 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
             error_propagate(errp, local_err);
             return;
         }
-    }
-
-    if (sphb->buid == (uint64_t)-1) {
-        error_setg(errp, "BUID not specified for PHB");
-        return;
-    }
-
-    if ((sphb->dma_liobn[0] == (uint32_t)-1) ||
-        ((sphb->dma_liobn[1] == (uint32_t)-1) && (windows_supported > 1))) {
-        error_setg(errp, "LIOBN(s) not specified for PHB");
-        return;
-    }
-
-    if (sphb->mem_win_addr == (hwaddr)-1) {
-        error_setg(errp, "Memory window address not specified for PHB");
-        return;
-    }
-
-    if (sphb->io_win_addr == (hwaddr)-1) {
-        error_setg(errp, "IO window address not specified for PHB");
+    } else {
+        error_setg(errp, "\"index\" for PAPR PHB is mandatory");
         return;
     }
 
     if (sphb->mem64_win_size != 0) {
-        if (sphb->mem64_win_addr == (hwaddr)-1) {
-            error_setg(errp,
-                       "64-bit memory window address not specified for PHB");
-            return;
-        }
-
         if (sphb->mem_win_size > SPAPR_PCI_MEM32_WIN_SIZE) {
             error_setg(errp, "32-bit memory window of size 0x%"HWADDR_PRIx
                        " (max 2 GiB)", sphb->mem_win_size);
@@ -1789,18 +1755,12 @@ static void spapr_phb_reset(DeviceState *qdev)
 
 static Property spapr_phb_properties[] = {
     DEFINE_PROP_UINT32("index", sPAPRPHBState, index, -1),
-    DEFINE_PROP_UINT64("buid", sPAPRPHBState, buid, -1),
-    DEFINE_PROP_UINT32("liobn", sPAPRPHBState, dma_liobn[0], -1),
-    DEFINE_PROP_UINT32("liobn64", sPAPRPHBState, dma_liobn[1], -1),
-    DEFINE_PROP_UINT64("mem_win_addr", sPAPRPHBState, mem_win_addr, -1),
     DEFINE_PROP_UINT64("mem_win_size", sPAPRPHBState, mem_win_size,
                        SPAPR_PCI_MEM32_WIN_SIZE),
-    DEFINE_PROP_UINT64("mem64_win_addr", sPAPRPHBState, mem64_win_addr, -1),
     DEFINE_PROP_UINT64("mem64_win_size", sPAPRPHBState, mem64_win_size,
                        SPAPR_PCI_MEM64_WIN_SIZE),
     DEFINE_PROP_UINT64("mem64_win_pciaddr", sPAPRPHBState, mem64_win_pciaddr,
                        -1),
-    DEFINE_PROP_UINT64("io_win_addr", sPAPRPHBState, io_win_addr, -1),
     DEFINE_PROP_UINT64("io_win_size", sPAPRPHBState, io_win_size,
                        SPAPR_PCI_IO_WIN_SIZE),
     DEFINE_PROP_BOOL("dynamic-reconfiguration", sPAPRPHBState, dr_enabled,
@@ -1937,6 +1897,13 @@ static const char *spapr_phb_root_bus_path(PCIHostState *host_bridge,
     return sphb->dtbusname;
 }
 
+static void spapr_phb_instance_init(Object *obj)
+{
+    sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(obj);
+
+    sphb->mem64_win_addr = (hwaddr)-1;
+}
+
 static void spapr_phb_class_init(ObjectClass *klass, void *data)
 {
     PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
@@ -1960,6 +1927,7 @@ static const TypeInfo spapr_phb_info = {
     .parent        = TYPE_PCI_HOST_BRIDGE,
     .instance_size = sizeof(sPAPRPHBState),
     .class_init    = spapr_phb_class_init,
+    .instance_init = spapr_phb_instance_init,
     .interfaces    = (InterfaceInfo[]) {
         { TYPE_HOTPLUG_HANDLER },
         { }


Re: [Qemu-devel] [RFC PATCH] spapr_pci: make index property mandatory
Posted by no-reply@patchew.org 6 years, 7 months ago
Hi,

This series seems to have some coding style problems. See output below for
more information:

Subject: [Qemu-devel] [RFC PATCH] spapr_pci: make index property mandatory
Message-id: 150512970330.16081.7553039475680302776.stgit@bahia
Type: series

=== TEST SCRIPT BEGIN ===
#!/bin/bash

BASE=base
n=1
total=$(git log --oneline $BASE.. | wc -l)
failed=0

git config --local diff.renamelimit 0
git config --local diff.renames True

commits="$(git log --format=%H --reverse $BASE..)"
for c in $commits; do
    echo "Checking PATCH $n/$total: $(git log -n 1 --format=%s $c)..."
    if ! git show $c --format=email | ./scripts/checkpatch.pl --mailback -; then
        failed=1
        echo
    fi
    n=$((n+1))
done

exit $failed
=== TEST SCRIPT END ===

Updating 3c8cf5a9c21ff8782164d1def7f44bd888713384
From https://github.com/patchew-project/qemu
 * [new tag]               patchew/150512970330.16081.7553039475680302776.stgit@bahia -> patchew/150512970330.16081.7553039475680302776.stgit@bahia
 t [tag update]            patchew/20170911110037.6567-1-lvivier@redhat.com -> patchew/20170911110037.6567-1-lvivier@redhat.com
Switched to a new branch 'test'
4379078966 spapr_pci: make index property mandatory

=== OUTPUT BEGIN ===
Checking PATCH 1/1: spapr_pci: make index property mandatory...
ERROR: spaces required around that '-' (ctx:VxV)
#125: FILE: hw/ppc/spapr_pci.c:1920:
+    sphb->mem64_win_addr = (hwaddr)-1;
                                    ^

total: 1 errors, 0 warnings, 92 lines checked

Your patch has style problems, please review.  If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.

=== OUTPUT END ===

Test command exited with code: 1


---
Email generated automatically by Patchew [http://patchew.org/].
Please send your feedback to patchew-devel@freelists.org
Re: [Qemu-devel] [RFC PATCH] spapr_pci: make index property mandatory
Posted by David Gibson 6 years, 7 months ago
On Mon, Sep 11, 2017 at 01:35:03PM +0200, Greg Kurz wrote:
> Creating several PHBs without index property confuses the DRC code
> and causes issues:
> - only the first index-less PHB is functional, the other ones will
>   silently ignore hotplugging of PCI devices
> - QEMU will even terminate if these PHBs have cold-plugged devices
> 
> qemu-system-ppc64: -device virtio-net,bus=pci2.0: an attached device
>  is still awaiting release
> 
> This happens because DR connectors for child PCI devices are created
> with a DRC index that is derived from the PHB's index property. If the
> PHBs are created without index, then the same value of -1 is used to
> compute the DRC indexes for both PHBs, hence causing the collision.
> 
> Also, the index property is used to compute the placement of the PHB's
> memory regions. It is limited to 31 or 255, depending on the machine
> type version. This fits well with the requirements of DRC indexes, which
> need the PHB index to be a 16-bit value.
> 
> This patch hence makes the index property mandatory. As a consequence,
> the PHB's memory regions and BUID are now always configured according
> to the index, and it is no longer possible to set them from the command
> line. This is a reasonable trade-off, as it is very unlikely that people
> create PHBs without index (at least libvirt doesn't do it FWIW).
> 
> We have to introduce a PHB instance init function to initialize the
> 64-bit window address to -1 because pseries-2.7 and older machines
> don't set it.
> 
> Signed-off-by: Greg Kurz <groug@kaod.org>
> ---
> 
> Hi,
> 
> This is a proposal to address the issue uncovered during the review of the
> PHB hotplug patches:
> 
> https://lists.nongnu.org/archive/html/qemu-ppc/2017-07/msg00557.html
> 
> I'd like to address this properly before resuming work on PHB hotplug
> itself. Please comment.

Yes, I think this looks good.  Commit message should explicitly
mention that this does break backwards compat - but we don't think the
non-index PHB feature was used in practice and the simplification is
worth it.

-- 
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