Skip to contentSkip to conversation

s390x/pci: Send correct event on hotplug.

PATCH Applied David Hildenbrand 4 participants 9 messages Jan 10, 2019 – Jan 15, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 s390x/pci: Send correct event on hotplug.

Comit 2c28c490571f ("s390x/pci: let pci devices start in configured mode") changed the initial state of zPCI devices from ZPCI_FS_STANDBY to ZPCI_FS_DISABLED (a.k.a. configured). However we still only send a HP_EVENT_RESERVED_TO_STANDBY event to the guest, indicating a wrong state.

Let's send a HP_EVENT_TO_CONFIGURED event instead, to match the actual state the device is in.

This fixes hotplugged devices having to be enabled explicitly in the guest e.g. via echo 1 > /sys/bus/pci/slots/00000000/power.

Fixes 2c28c490571f ("s390x/pci: let pci devices start in configured mode")
---
 hw/s390x/s390-pci-bus.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/s390x/s390-pci-bus.c +1 -1
index 15759b6514..7f911b216a 100644
@@ -899,7 +899,7 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
         }
 
         if (dev->hotplugged) {
-            s390_pci_generate_plug_event(HP_EVENT_RESERVED_TO_STANDBY,
+            s390_pci_generate_plug_event(HP_EVENT_TO_CONFIGURED ,
                                          pbdev->fh, pbdev->fid);
         }
     } else if (object_dynamic_cast(OBJECT(dev), TYPE_S390_PCI_DEVICE)) {
Signature · 1 line
2.17.2
#1 lore raw
Quoted 6 lines · David Hildenbrand

Comit 2c28c490571f ("s390x/pci: let pci devices start in configured mode") changed the initial state of zPCI devices from ZPCI_FS_STANDBY to ZPCI_FS_DISABLED (a.k.a. configured). However we still only send a HP_EVENT_RESERVED_TO_STANDBY event to the guest, indicating a wrong state.

Quoted 8 lines

Let's send a HP_EVENT_TO_CONFIGURED event instead, to match the actual state the device is in.

This fixes hotplugged devices having to be enabled explicitly in the guest e.g. via echo 1 > /sys/bus/pci/slots/00000000/power.

Fixes 2c28c490571f ("s390x/pci: let pci devices start in configured mode")

If this patch is the right thing to do, then

1. s/Report-by/Reported-by/
2. Dropping the "." from the subject

(yes, it was late)

I wonder if we should do both events sequentially, but as I don't have access to the architecture I have to rely on that this works :)

Quoted 19 lines
---
 hw/s390x/s390-pci-bus.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/s390x/s390-pci-bus.c +1 -1
index 15759b6514..7f911b216a 100644
@@ -899,7 +899,7 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
         }
 
         if (dev->hotplugged) {
-            s390_pci_generate_plug_event(HP_EVENT_RESERVED_TO_STANDBY,
+            s390_pci_generate_plug_event(HP_EVENT_TO_CONFIGURED ,
                                          pbdev->fh, pbdev->fid);
         }
     } else if (object_dynamic_cast(OBJECT(dev), TYPE_S390_PCI_DEVICE)) {
Signature · 3 lines
Thanks,

David / dhildenb
Tested-by
#2 lore raw
Quoted 7 lines · On Fri, 11 Jan 2019 08:16:41 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 6 lines · David Hildenbrand

Comit 2c28c490571f ("s390x/pci: let pci devices start in configured mode") changed the initial state of zPCI devices from ZPCI_FS_STANDBY to ZPCI_FS_DISABLED (a.k.a. configured). However we still only send a HP_EVENT_RESERVED_TO_STANDBY event to the guest, indicating a wrong state.

Quoted 8 lines
Quoted 8 lines

Let's send a HP_EVENT_TO_CONFIGURED event instead, to match the actual state the device is in.

This fixes hotplugged devices having to be enabled explicitly in the guest e.g. via echo 1 > /sys/bus/pci/slots/00000000/power.

Fixes 2c28c490571f ("s390x/pci: let pci devices start in configured mode")

Cool, works for me as well.

Do we want to cc:stable? Probably not, as it's more annoying than critical, and pci hotplug does not seem to be much used on s390x.

Quoted 7 lines

If this patch is the right thing to do, then

1. s/Report-by/Reported-by/
2. Dropping the "." from the subject

(yes, it was late)

:) Can do while applying.

Quoted 3 lines

I wonder if we should do both events sequentially, but as I don't have access to the architecture I have to rely on that this works :)

Yep, let's wait for feedback from folks with architecture access.

Quoted 22 lines
Quoted 20 lines
---
 hw/s390x/s390-pci-bus.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/s390x/s390-pci-bus.c +1 -1
index 15759b6514..7f911b216a 100644
@@ -899,7 +899,7 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
         }
 
         if (dev->hotplugged) {
-            s390_pci_generate_plug_event(HP_EVENT_RESERVED_TO_STANDBY,
+            s390_pci_generate_plug_event(HP_EVENT_TO_CONFIGURED ,
                                          pbdev->fh, pbdev->fid);
         }
     } else if (object_dynamic_cast(OBJECT(dev), TYPE_S390_PCI_DEVICE)) {
Reviewed-by
#3 lore raw
Quoted 34 lines · Cornelia Huck
Quoted 7 lines · On Fri, 11 Jan 2019 08:16:41 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 6 lines · David Hildenbrand

Comit 2c28c490571f ("s390x/pci: let pci devices start in configured mode") changed the initial state of zPCI devices from ZPCI_FS_STANDBY to ZPCI_FS_DISABLED (a.k.a. configured). However we still only send a HP_EVENT_RESERVED_TO_STANDBY event to the guest, indicating a wrong state.

Quoted 8 lines
Quoted 8 lines

Let's send a HP_EVENT_TO_CONFIGURED event instead, to match the actual state the device is in.

This fixes hotplugged devices having to be enabled explicitly in the guest e.g. via echo 1 > /sys/bus/pci/slots/00000000/power.

Fixes 2c28c490571f ("s390x/pci: let pci devices start in configured mode")

Cool, works for me as well.

Do we want to cc:stable? Probably not, as it's more annoying than critical, and pci hotplug does not seem to be much used on s390x.

Quoted 7 lines

If this patch is the right thing to do, then

1. s/Report-by/Reported-by/
2. Dropping the "." from the subject

(yes, it was late)

Quoted 8 lines

:) Can do while applying.

Quoted 3 lines

I wonder if we should do both events sequentially, but as I don't have access to the architecture I have to rely on that this works :)

Yep, let's wait for feedback from folks with architecture access.

Works fine on the architecture too.

Seems the logical thing to do for me.

Quoted 24 lines
Quoted 22 lines
Quoted 20 lines
---
  hw/s390x/s390-pci-bus.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
hw/s390x/s390-pci-bus.c +1 -1
index 15759b6514..7f911b216a 100644
@@ -899,7 +899,7 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
          }
 
          if (dev->hotplugged) {
-            s390_pci_generate_plug_event(HP_EVENT_RESERVED_TO_STANDBY,
+            s390_pci_generate_plug_event(HP_EVENT_TO_CONFIGURED ,
                                           pbdev->fh, pbdev->fid);
          }
      } else if (object_dynamic_cast(OBJECT(dev), TYPE_S390_PCI_DEVICE)) {
Signature · 2 lines
Pierre Morel
Linux/KVM/QEMU in Böblingen - Germany
#4 lore raw

[restored cc:s]

Quoted 41 lines · On Mon, 14 Jan 2019 11:06:19 +0100 Pierre Morel <[email protected]> wrote:
Quoted 40 lines · Cornelia Huck
Quoted 7 lines · On Fri, 11 Jan 2019 08:16:41 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 6 lines · David Hildenbrand

Comit 2c28c490571f ("s390x/pci: let pci devices start in configured mode") changed the initial state of zPCI devices from ZPCI_FS_STANDBY to ZPCI_FS_DISABLED (a.k.a. configured). However we still only send a HP_EVENT_RESERVED_TO_STANDBY event to the guest, indicating a wrong state.

Quoted 8 lines
Quoted 8 lines

Let's send a HP_EVENT_TO_CONFIGURED event instead, to match the actual state the device is in.

This fixes hotplugged devices having to be enabled explicitly in the guest e.g. via echo 1 > /sys/bus/pci/slots/00000000/power.

Fixes 2c28c490571f ("s390x/pci: let pci devices start in configured mode")

Cool, works for me as well.

Do we want to cc:stable? Probably not, as it's more annoying than critical, and pci hotplug does not seem to be much used on s390x.

Quoted 7 lines

If this patch is the right thing to do, then

1. s/Report-by/Reported-by/
2. Dropping the "." from the subject

(yes, it was late)

:) Can do while applying.

Quoted 3 lines

I wonder if we should do both events sequentially, but as I don't have access to the architecture I have to rely on that this works :)

Quoted 8 lines
Quoted 2 lines

Yep, let's wait for feedback from folks with architecture access.

Works fine on the architecture too.

Seems the logical thing to do for me.

Thanks for checking.

I'd like to queue this, but I'd like an ack from Collin as well.

Quoted 28 lines
Quoted 26 lines
Quoted 24 lines
Quoted 22 lines
---
  hw/s390x/s390-pci-bus.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
hw/s390x/s390-pci-bus.c +1 -1
index 15759b6514..7f911b216a 100644
@@ -899,7 +899,7 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
          }
 
          if (dev->hotplugged) {
-            s390_pci_generate_plug_event(HP_EVENT_RESERVED_TO_STANDBY,
+            s390_pci_generate_plug_event(HP_EVENT_TO_CONFIGURED ,
                                           pbdev->fh, pbdev->fid);
          }
      } else if (object_dynamic_cast(OBJECT(dev), TYPE_S390_PCI_DEVICE)) {
Reviewed-by
#5 lore raw
Quoted 51 lines · Cornelia Huck

[restored cc:s]

Quoted 46 lines · On Mon, 14 Jan 2019 11:06:19 +0100 Pierre Morel <[email protected]> wrote:
Quoted 34 lines · Cornelia Huck
Quoted 7 lines · On Fri, 11 Jan 2019 08:16:41 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 6 lines · David Hildenbrand

Comit 2c28c490571f ("s390x/pci: let pci devices start in configured mode") changed the initial state of zPCI devices from ZPCI_FS_STANDBY to ZPCI_FS_DISABLED (a.k.a. configured). However we still only send a HP_EVENT_RESERVED_TO_STANDBY event to the guest, indicating a wrong state.

Quoted 8 lines
Quoted 8 lines

Let's send a HP_EVENT_TO_CONFIGURED event instead, to match the actual state the device is in.

This fixes hotplugged devices having to be enabled explicitly in the guest e.g. via echo 1 > /sys/bus/pci/slots/00000000/power.

Fixes 2c28c490571f ("s390x/pci: let pci devices start in configured mode")

Cool, works for me as well.

Do we want to cc:stable? Probably not, as it's more annoying than critical, and pci hotplug does not seem to be much used on s390x.

Quoted 7 lines

If this patch is the right thing to do, then

1. s/Report-by/Reported-by/
2. Dropping the "." from the subject

(yes, it was late)

Quoted 8 lines

:) Can do while applying.

Quoted 3 lines

I wonder if we should do both events sequentially, but as I don't have access to the architecture I have to rely on that this works :)

Yep, let's wait for feedback from folks with architecture access.

Works fine on the architecture too.

Quoted 8 lines
Quoted 3 lines

Seems the logical thing to do for me.

Thanks for checking.

I'd like to queue this, but I'd like an ack from Collin as well.

Would you mind adding a comment somewhere that states something like "we can safely bypass the standby state when PCI hotplugging for a guest" just to be clear that QEMU is a bit different from how we handle it on the LPAR level?

That comment would more-or-less clarify why we set the ZPCI_FS_<STATE> directly to disabled instead of to standby when hotplugging (which, AFAIU, is the order how things occur at the LPAR level)

Otherwise,

Quoted 30 lines
Quoted 28 lines
Quoted 26 lines
Quoted 24 lines
Quoted 22 lines
---
  hw/s390x/s390-pci-bus.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
hw/s390x/s390-pci-bus.c +1 -1
index 15759b6514..7f911b216a 100644
@@ -899,7 +899,7 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
          }
 
          if (dev->hotplugged) {
-            s390_pci_generate_plug_event(HP_EVENT_RESERVED_TO_STANDBY,
+            s390_pci_generate_plug_event(HP_EVENT_TO_CONFIGURED ,
                                           pbdev->fh, pbdev->fid);
          }
      } else if (object_dynamic_cast(OBJECT(dev), TYPE_S390_PCI_DEVICE)) {
Signature · 2 lines
Respectfully,
- Collin Walling
#6 lore raw
Quoted 61 lines · Collin Walling
Quoted 51 lines · Cornelia Huck

[restored cc:s]

Quoted 46 lines · On Mon, 14 Jan 2019 11:06:19 +0100 Pierre Morel <[email protected]> wrote:
Quoted 34 lines · Cornelia Huck
Quoted 7 lines · On Fri, 11 Jan 2019 08:16:41 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 6 lines · David Hildenbrand

Comit 2c28c490571f ("s390x/pci: let pci devices start in configured mode") changed the initial state of zPCI devices from ZPCI_FS_STANDBY to ZPCI_FS_DISABLED (a.k.a. configured). However we still only send a HP_EVENT_RESERVED_TO_STANDBY event to the guest, indicating a wrong state.

Quoted 8 lines
Quoted 8 lines

Let's send a HP_EVENT_TO_CONFIGURED event instead, to match the actual state the device is in.

This fixes hotplugged devices having to be enabled explicitly in the guest e.g. via echo 1 > /sys/bus/pci/slots/00000000/power.

Fixes 2c28c490571f ("s390x/pci: let pci devices start in configured mode")

Cool, works for me as well.

Do we want to cc:stable? Probably not, as it's more annoying than critical, and pci hotplug does not seem to be much used on s390x.

Quoted 7 lines

If this patch is the right thing to do, then

1. s/Report-by/Reported-by/
2. Dropping the "." from the subject

(yes, it was late)

Quoted 8 lines

:) Can do while applying.

Quoted 3 lines

I wonder if we should do both events sequentially, but as I don't have access to the architecture I have to rely on that this works :)

Yep, let's wait for feedback from folks with architecture access.

Works fine on the architecture too.

Quoted 8 lines
Quoted 3 lines

Seems the logical thing to do for me.

Thanks for checking.

I'd like to queue this, but I'd like an ack from Collin as well.

Quoted 8 lines

Would you mind adding a comment somewhere that states something like "we can safely bypass the standby state when PCI hotplugging for a guest" just to be clear that QEMU is a bit different from how we handle it on the LPAR level?

That comment would more-or-less clarify why we set the ZPCI_FS_<STATE> directly to disabled instead of to standby when hotplugging (which, AFAIU, is the order how things occur at the LPAR level)

This patch relies on Christians patch, where the general concept was explained. As we changed the initial state, we have to send a corresponding hotplug event. But still we can add a comment to shine some light on the general concept.

@Conny, can you add after the first paragraph: (or let me know if you want a respin)

"On real HW, a PCI device always pops up in the STANDBY state. In QEMU, we decided to let it show up directly in the configured state (as configuring it is otherwise just an extra burden for the admin). We can safely bypass the STANDBY state when hotplugging PCI devices to a guest."

Quoted 4 lines

Otherwise,

Thanks!

Signature · 3 lines
Thanks,

David / dhildenb
#7 lore raw
Quoted 76 lines · On Mon, 14 Jan 2019 21:59:58 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 61 lines · Collin Walling
Quoted 51 lines · Cornelia Huck

[restored cc:s]

Quoted 46 lines · On Mon, 14 Jan 2019 11:06:19 +0100 Pierre Morel <[email protected]> wrote:
Quoted 34 lines · Cornelia Huck
Quoted 7 lines · On Fri, 11 Jan 2019 08:16:41 +0100 David Hildenbrand <[email protected]> wrote:
Quoted 6 lines · David Hildenbrand

Comit 2c28c490571f ("s390x/pci: let pci devices start in configured mode") changed the initial state of zPCI devices from ZPCI_FS_STANDBY to ZPCI_FS_DISABLED (a.k.a. configured). However we still only send a HP_EVENT_RESERVED_TO_STANDBY event to the guest, indicating a wrong state.

Quoted 8 lines
Quoted 8 lines

Let's send a HP_EVENT_TO_CONFIGURED event instead, to match the actual state the device is in.

This fixes hotplugged devices having to be enabled explicitly in the guest e.g. via echo 1 > /sys/bus/pci/slots/00000000/power.

Fixes 2c28c490571f ("s390x/pci: let pci devices start in configured mode")

Cool, works for me as well.

Do we want to cc:stable? Probably not, as it's more annoying than critical, and pci hotplug does not seem to be much used on s390x.

Quoted 7 lines

If this patch is the right thing to do, then

1. s/Report-by/Reported-by/
2. Dropping the "." from the subject

(yes, it was late)

Quoted 8 lines

:) Can do while applying.

Quoted 3 lines

I wonder if we should do both events sequentially, but as I don't have access to the architecture I have to rely on that this works :)

Yep, let's wait for feedback from folks with architecture access.

Works fine on the architecture too.

Quoted 8 lines
Quoted 3 lines

Seems the logical thing to do for me.

Thanks for checking.

I'd like to queue this, but I'd like an ack from Collin as well.

Quoted 8 lines

Would you mind adding a comment somewhere that states something like "we can safely bypass the standby state when PCI hotplugging for a guest" just to be clear that QEMU is a bit different from how we handle it on the LPAR level?

That comment would more-or-less clarify why we set the ZPCI_FS_<STATE> directly to disabled instead of to standby when hotplugging (which, AFAIU, is the order how things occur at the LPAR level)

This patch relies on Christians patch, where the general concept was explained. As we changed the initial state, we have to send a corresponding hotplug event. But still we can add a comment to shine some light on the general concept.

Quoted 7 lines

@Conny, can you add after the first paragraph: (or let me know if you want a respin)

"On real HW, a PCI device always pops up in the STANDBY state. In QEMU, we decided to let it show up directly in the configured state (as configuring it is otherwise just an extra burden for the admin). We can safely bypass the STANDBY state when hotplugging PCI devices to a guest."

I'll just add that text.

Quoted 8 lines
Quoted 5 lines

Otherwise,

Thanks!

Thanks!

#8 lore raw
In reply to David Hildenbrand
Quoted 12 lines · On Thu, 10 Jan 2019 22:03:58 +0100 David Hildenbrand <[email protected]> wrote:

Comit 2c28c490571f ("s390x/pci: let pci devices start in configured mode") changed the initial state of zPCI devices from ZPCI_FS_STANDBY to ZPCI_FS_DISABLED (a.k.a. configured). However we still only send a HP_EVENT_RESERVED_TO_STANDBY event to the guest, indicating a wrong state.

Let's send a HP_EVENT_TO_CONFIGURED event instead, to match the actual state the device is in.

This fixes hotplugged devices having to be enabled explicitly in the guest e.g. via echo 1 > /sys/bus/pci/slots/00000000/power.

Quoted 6 lines
Fixes 2c28c490571f ("s390x/pci: let pci devices start in configured mode")
---
 hw/s390x/s390-pci-bus.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Thanks, applied.