Skip to contentSkip to conversation

vfio-ccw: Enable transparent CCW IPL from DASD

PATCH Superseded Jared Rossi 2 participants 6 messages Apr 17, 2020 – Apr 21, 2020

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

cover vfio-ccw: Enable transparent CCW IPL from DASD

Remove the explicit prefetch check when using vfio-ccw devices. This check is not needed as all Linux channel programs are intended to use prefetch and will be executed in the same way regardless.

Jared Rossi (1):
vfio-ccw: Enable transparent CCW IPL from DASD

 hw/vfio/ccw.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
Signature · 1 line
2.21.1
#1 lore raw

PATCH 1/1 vfio-ccw: Enable transparent CCW IPL from DASD

Remove the explicit prefetch check when using vfio-ccw devices. This check is not needed as all Linux channel programs are intended to use prefetch and will be executed in the same way regardless.

---
 hw/vfio/ccw.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
hw/vfio/ccw.c +3 -10
index 50cc2ec75c..e649377b68 100644
@@ -74,16 +74,9 @@ static IOInstEnding vfio_ccw_handle_request(SubchDev *sch)
     struct ccw_io_region *region = vcdev->io_region;
     int ret;
 
-    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH)) {
-        if (!(vcdev->force_orb_pfch)) {
-            warn_once_pfch(vcdev, sch, "requires PFCH flag set");
-            sch_gen_unit_exception(sch);
-            css_inject_io_interrupt(sch);
-            return IOINST_CC_EXPECTED;
-        } else {
-            sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
-            warn_once_pfch(vcdev, sch, "PFCH flag forced");
-        }
+    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH) && vcdev->force_orb_pfch) {
+        sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
+        warn_once_pfch(vcdev, sch, "PFCH flag forced");
     }
 
     QEMU_BUILD_BUG_ON(sizeof(region->orb_area) != sizeof(ORB));
Signature · 1 line
2.21.1
#2 lore raw
Quoted 3 lines · On Fri, 17 Apr 2020 14:38:38 -0400 Jared Rossi <[email protected]> wrote:

Remove the explicit prefetch check when using vfio-ccw devices. This check is not needed as all Linux channel programs are intended to use prefetch and will be executed in the same way regardless.

As already commented on the Linux patch: Can we log something, so this is debuggable if this statement does not hold true in the future?

Quoted 18 lines
---
 hw/vfio/ccw.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
hw/vfio/ccw.c +0 -4
index 50cc2ec75c..e649377b68 100644
@@ -74,16 +74,9 @@ static IOInstEnding vfio_ccw_handle_request(SubchDev *sch)
     struct ccw_io_region *region = vcdev->io_region;
     int ret;
 
-    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH)) {
-        if (!(vcdev->force_orb_pfch)) {
-            warn_once_pfch(vcdev, sch, "requires PFCH flag set");
-            sch_gen_unit_exception(sch);

hw/vfio/ccw.c static IOInstEnding vfio_ccw_handle_request(SubchDev *sch) · line 80

Quoted 10 lines
hw/vfio/ccw.c +3 -6
-            css_inject_io_interrupt(sch);
-            return IOINST_CC_EXPECTED;
-        } else {
-            sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
-            warn_once_pfch(vcdev, sch, "PFCH flag forced");
-        }
+    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH) && vcdev->force_orb_pfch) {
+        sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
+        warn_once_pfch(vcdev, sch, "PFCH flag forced");
     }

What happens when you run it with an old kernel? I guess the I/O is only rejected later (after a trip into the kernel), but has that path ever been tested?

Quoted 2 lines
hw/vfio/ccw.c +0 -0
     QEMU_BUILD_BUG_ON(sizeof(region->orb_area) != sizeof(ORB));
#3 lore raw
Quoted 34 lines · On Mon, 20 Apr 2020 14:26:17 +0200 Cornelia Huck <[email protected]> wrote:
Quoted 3 lines · On Fri, 17 Apr 2020 14:38:38 -0400 Jared Rossi <[email protected]> wrote:

Remove the explicit prefetch check when using vfio-ccw devices. This check is not needed as all Linux channel programs are intended to use prefetch and will be executed in the same way regardless.

As already commented on the Linux patch: Can we log something, so this is debuggable if this statement does not hold true in the future?

Quoted 24 lines
---
 hw/vfio/ccw.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
hw/vfio/ccw.c +0 -10
index 50cc2ec75c..e649377b68 100644
@@ -74,16 +74,9 @@ static IOInstEnding vfio_ccw_handle_request(SubchDev *sch)
     struct ccw_io_region *region = vcdev->io_region;
     int ret;
 
-    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH)) {
-        if (!(vcdev->force_orb_pfch)) {
-            warn_once_pfch(vcdev, sch, "requires PFCH flag set");
-            sch_gen_unit_exception(sch);
-            css_inject_io_interrupt(sch);
-            return IOINST_CC_EXPECTED;
-        } else {
-            sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
-            warn_once_pfch(vcdev, sch, "PFCH flag forced");
-        }
Quoted 12 lines

hw/vfio/ccw.c static IOInstEnding vfio_ccw_handle_request(SubchDev *sch) · line 80

Quoted 4 lines
hw/vfio/ccw.c +3 -0
+    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH) && vcdev->force_orb_pfch) {
+        sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
+        warn_once_pfch(vcdev, sch, "PFCH flag forced");
     }  

What happens when you run it with an old kernel? I guess the I/O is only rejected later (after a trip into the kernel), but has that path ever been tested?

Quoted 2 lines
hw/vfio/ccw.c +0 -0
     QEMU_BUILD_BUG_ON(sizeof(region->orb_area) != sizeof(ORB));  

Oh, and do we want to deprecate the force prefetch interface in the future? We probably need to wait a bit, until the kernel changes have become widely available.

#4 lore raw
Quoted 6 lines · Cornelia Huck
Quoted 3 lines · On Mon, 20 Apr 2020 14:26:17 +0200 Cornelia Huck <[email protected]> wrote:

On Fri, 17 Apr 2020 14:38:38 -0400
Jared Rossi <[email protected]> wrote:

Quoted 7 lines
Quoted 7 lines
Quoted 3 lines

Remove the explicit prefetch check when using vfio-ccw devices. This check is not needed as all Linux channel programs are intended to use prefetch and will be executed in the same way regardless.

As already commented on the Linux patch: Can we log something, so this is debuggable if this statement does not hold true in the future?

Agreed. I will work on debugging improvements so that any future issues related to unintended prefetching are more clearly logged.

Quoted 25 lines
Quoted 25 lines
Quoted 25 lines
---
 hw/vfio/ccw.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
hw/vfio/ccw.c +1 -10
index 50cc2ec75c..e649377b68 100644
@@ -74,16 +74,9 @@ static IOInstEnding vfio_ccw_handle_request(SubchDev *sch)
     struct ccw_io_region *region = vcdev->io_region;
     int ret;
 
-    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH)) {
-        if (!(vcdev->force_orb_pfch)) {
-            warn_once_pfch(vcdev, sch, "requires PFCH flag set");
-            sch_gen_unit_exception(sch);
-            css_inject_io_interrupt(sch);
-            return IOINST_CC_EXPECTED;
-        } else {
-            sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
-            warn_once_pfch(vcdev, sch, "PFCH flag forced");
-        }
+    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH) && vcdev->force_orb_pfch) {
Quoted 8 lines
Quoted 8 lines

hw/vfio/ccw.c static IOInstEnding vfio_ccw_handle_request(SubchDev *sch) · line 80

Quoted 3 lines
hw/vfio/ccw.c +2 -0
+        sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
+        warn_once_pfch(vcdev, sch, "PFCH flag forced");
     }

What happens when you run it with an old kernel? I guess the I/O is only rejected later (after a trip into the kernel), but has that path ever been tested?

Yes, this was tested and you are correct that the kernel will reject the I/O unless the corresponding patch is also applied there. I will revisit this path while I'm updating the logging to ensure that any potential interactions are appropriately considered.

Quoted 7 lines
Quoted 3 lines

hw/vfio/ccw.c static IOInstEnding vfio_ccw_handle_request(SubchDev *sch) · line 82

Quoted 2 lines
hw/vfio/ccw.c +0 -0
     QEMU_BUILD_BUG_ON(sizeof(region->orb_area) != sizeof(ORB));

Oh, and do we want to deprecate the force prefetch interface in the future? We probably need to wait a bit, until the kernel changes have become widely available.

Yes, I think we will want to deprecate it at an appropriate time in the future.

#5 lore raw
Quoted 11 lines · On Mon, 20 Apr 2020 18:35:58 -0400 Jared Rossi <[email protected]> wrote:
Quoted 10 lines · Cornelia Huck
Quoted 7 lines · On Mon, 20 Apr 2020 14:26:17 +0200 Cornelia Huck <[email protected]> wrote:
Quoted 3 lines · On Fri, 17 Apr 2020 14:38:38 -0400 Jared Rossi <[email protected]> wrote:

Remove the explicit prefetch check when using vfio-ccw devices. This check is not needed as all Linux channel programs are intended to use prefetch and will be executed in the same way regardless.

Quoted 6 lines
Quoted 3 lines
Quoted 3 lines

As already commented on the Linux patch: Can we log something, so this is debuggable if this statement does not hold true in the future?

Agreed. I will work on debugging improvements so that any future issues related to unintended prefetching are more clearly logged.

Great.

Quoted 35 lines
Quoted 26 lines
Quoted 26 lines
Quoted 26 lines
---
 hw/vfio/ccw.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)
hw/vfio/ccw.c +1 -10
index 50cc2ec75c..e649377b68 100644
@@ -74,16 +74,9 @@ static IOInstEnding vfio_ccw_handle_request(SubchDev *sch)
     struct ccw_io_region *region = vcdev->io_region;
     int ret;
 
-    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH)) {
-        if (!(vcdev->force_orb_pfch)) {
-            warn_once_pfch(vcdev, sch, "requires PFCH flag set");
-            sch_gen_unit_exception(sch);
-            css_inject_io_interrupt(sch);
-            return IOINST_CC_EXPECTED;
-        } else {
-            sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
-            warn_once_pfch(vcdev, sch, "PFCH flag forced");
-        }
+    if (!(sch->orb.ctrl0 & ORB_CTRL0_MASK_PFCH) && vcdev->force_orb_pfch) {
Quoted 8 lines
Quoted 8 lines

hw/vfio/ccw.c static IOInstEnding vfio_ccw_handle_request(SubchDev *sch) · line 80

Quoted 3 lines
hw/vfio/ccw.c +2 -0
+        sch->orb.ctrl0 |= ORB_CTRL0_MASK_PFCH;
+        warn_once_pfch(vcdev, sch, "PFCH flag forced");
     }  

What happens when you run it with an old kernel? I guess the I/O is only rejected later (after a trip into the kernel), but has that path ever been tested?

Quoted 7 lines

Yes, this was tested and you are correct that the kernel will reject the I/O unless
the corresponding patch is also applied there. I will revisit this path while I'm
updating the logging to ensure that any potential interactions are appropriately
considered.

I've looked at the code again and it seems that the kernel will end up signaling -EOPNOTSUPP to us on that case, which causes the same unit exception as without this patch, so we should be all good.

Quoted 12 lines
Quoted 8 lines
Quoted 4 lines

hw/vfio/ccw.c static IOInstEnding vfio_ccw_handle_request(SubchDev *sch) · line 82

Quoted 3 lines
hw/vfio/ccw.c +0 -0
     QEMU_BUILD_BUG_ON(sizeof(region->orb_area) != sizeof(ORB));  

Oh, and do we want to deprecate the force prefetch interface in the future? We probably need to wait a bit, until the kernel changes have become widely available.

Yes, I think we will want to deprecate it at an appropriate time in the future.