QEMU-Devel Archive on lore.kernel.org
 help / color / Atom feed
From: Mahesh Salgaonkar <mahesh@linux.ibm.com>
To: Qemu-devel <qemu-devel@nongnu.org>
Cc: Alexey Kardashevskiy <aik@ozlabs.ru>,
	Oliver O'Halloran <oohall@gmail.com>,
	Daniel Henrique Barboza <danielhb413@gmail.com>,
	Qemu-ppc <qemu-ppc@nongnu.org>
Subject: [PATCH updated v2] spapr: Fix EEH capability issue on KVM guest for PCI passthru
Date: Wed, 05 May 2021 20:18:27 +0530
Message-ID: <162022601665.118720.1424074431457537864.stgit@jupiter> (raw)

With upstream kernel, especially after commit 98ba956f6a389
("powerpc/pseries/eeh: Rework device EEH PE determination") we see that KVM
guest isn't able to enable EEH option for PCI pass-through devices anymore.

[root@atest-guest ~]# dmesg | grep EEH
[    0.032337] EEH: pSeries platform initialized
[    0.298207] EEH: No capable adapters found: recovery disabled.
[root@atest-guest ~]#

So far the linux kernel was assuming pe_config_addr equal to device's
config_addr and using it to enable EEH on the PE through ibm,set-eeh-option
RTAS call. Which wasn't the correct way as per PAPR. The linux kernel
commit 98ba956f6a389 fixed this flow. With that fixed, linux now uses PE
config address returned by ibm,get-config-addr-info2 RTAS call to enable
EEH option per-PE basis instead of per-device basis. However this has
uncovered a bug in qemu where ibm,set-eeh-option is treating PE config
address as per-device config address.

Hence in qemu guest with recent kernel the ibm,set-eeh-option RTAS call
fails with -3 return value indicating that there is no PCI device exist for
the specified PE config address. The rtas_ibm_set_eeh_option call uses
pci_find_device() to get the PC device that matches specific bus and devfn
extracted from PE config address passed as argument. Thus it tries to map
the PE config address to a single specific PCI device 'bus->devices[devfn]'
which always results into checking device on slot 0 'bus->devices[0]'.
This succeeds when there is a pass-through device (vfio-pci) present in
slot 0. But in cases where there is no pass-through device present in slot
0, but present in non-zero slots, ibm,set-eeh-option call fails to enable
the EEH capability.

hw/ppc/spapr_pci_vfio.c: spapr_phb_vfio_eeh_set_option()
   case RTAS_EEH_ENABLE: {
        PCIHostState *phb;
        PCIDevice *pdev;

        /*
         * The EEH functionality is enabled on basis of PCI device,
         * instead of PE. We need check the validity of the PCI
         * device address.
         */
        phb = PCI_HOST_BRIDGE(sphb);
        pdev = pci_find_device(phb->bus,
                               (addr >> 16) & 0xFF, (addr >> 8) & 0xFF);
        if (!pdev || !object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
            return RTAS_OUT_PARAM_ERROR;
        }

hw/pci/pci.c:pci_find_device()

PCIDevice *pci_find_device(PCIBus *bus, int bus_num, uint8_t devfn)
{
    bus = pci_find_bus_nr(bus, bus_num);

    if (!bus)
        return NULL;

    return bus->devices[devfn];
}

This patch fixes ibm,set-eeh-option to check for presence of any PCI device
(vfio-pci) under specified bus and enable the EEH if found. The current
code already makes sure that all the devices on that bus are from same
iommu group (within same PE) and fail very early if it does not.

After this fix guest is able to find EEH capable devices and enable EEH
recovery on it.

[root@atest-guest ~]# dmesg | grep EEH
[    0.048139] EEH: pSeries platform initialized
[    0.405115] EEH: Capable adapter found: recovery enabled.
[root@atest-guest ~]#

Signed-off-by: Mahesh Salgaonkar <mahesh@linux.ibm.com>
---
Change in v2:
- Fix ibm,set-eeh-option instead of returning per-device PE config address.
- Changed patch subject line.
---
 hw/ppc/spapr_pci_vfio.c |   27 ++++++++++++++++++++-------
 1 file changed, 20 insertions(+), 7 deletions(-)

diff --git a/hw/ppc/spapr_pci_vfio.c b/hw/ppc/spapr_pci_vfio.c
index e0547b1740..b30020da6a 100644
--- a/hw/ppc/spapr_pci_vfio.c
+++ b/hw/ppc/spapr_pci_vfio.c
@@ -47,6 +47,16 @@ void spapr_phb_vfio_reset(DeviceState *qdev)
     spapr_phb_vfio_eeh_reenable(SPAPR_PCI_HOST_BRIDGE(qdev));
 }
 
+static void spapr_eeh_pci_find_device(PCIBus *bus, PCIDevice *pdev,
+                                      void *opaque)
+{
+    bool *found = opaque;
+
+    if (object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
+        *found = true;
+    }
+}
+
 int spapr_phb_vfio_eeh_set_option(SpaprPhbState *sphb,
                                   unsigned int addr, int option)
 {
@@ -59,17 +69,20 @@ int spapr_phb_vfio_eeh_set_option(SpaprPhbState *sphb,
         break;
     case RTAS_EEH_ENABLE: {
         PCIHostState *phb;
-        PCIDevice *pdev;
+        bool found = false;
 
         /*
-         * The EEH functionality is enabled on basis of PCI device,
-         * instead of PE. We need check the validity of the PCI
-         * device address.
+         * The EEH functionality is enabled per sphb level instead of
+         * per PCI device. We just need to check the validity of the PCI
+         * pass-through devices (vfio-pci) under this sphb bus.
+         * We have already validated that all the devices under this sphb
+         * are from same iommu group (within same PE) before comming here.
          */
         phb = PCI_HOST_BRIDGE(sphb);
-        pdev = pci_find_device(phb->bus,
-                               (addr >> 16) & 0xFF, (addr >> 8) & 0xFF);
-        if (!pdev || !object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
+        pci_for_each_device(phb->bus, (addr >> 16) & 0xFF,
+                            spapr_eeh_pci_find_device, &found);
+
+        if (!found) {
             return RTAS_OUT_PARAM_ERROR;
         }
 




             reply index

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-05-05 14:48 Mahesh Salgaonkar [this message]
2021-05-10 20:03 ` Daniel Henrique Barboza
2021-05-13  3:08 ` David Gibson
2021-05-14  2:03   ` Oliver O'Halloran
2021-05-17  6:36     ` 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=162022601665.118720.1424074431457537864.stgit@jupiter \
    --to=mahesh@linux.ibm.com \
    --cc=aik@ozlabs.ru \
    --cc=danielhb413@gmail.com \
    --cc=oohall@gmail.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

QEMU-Devel Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/qemu-devel/0 qemu-devel/git/0.git
	git clone --mirror https://lore.kernel.org/qemu-devel/1 qemu-devel/git/1.git
	git clone --mirror https://lore.kernel.org/qemu-devel/2 qemu-devel/git/2.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 qemu-devel qemu-devel/ https://lore.kernel.org/qemu-devel \
		qemu-devel@nongnu.org
	public-inbox-index qemu-devel

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.nongnu.qemu-devel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git