Skip to contentSkip to conversation

vfio: blacklist loading of unstable roms

PATCH Superseded Bandan Das 3 participants 16 messages Feb 19, 2014 – Feb 24, 2014

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Bandan Das

author
#0 lore raw

v2 cover vfio: blacklist loading of unstable roms

v2:
1. Break up into two patches separating the infrastructural changes
2. Change vendor_id and device_id type to uint16_t
3. Rename struct for blacklisted devid and vendorid to more meaningful names
4. Remove unnecessary rom_quirk variable and just call vfio_blacklist_opt_rom() in vfio_pci_size_rom()

Bandan Das (2):
pci: change default value of rom_bar to 2
vfio: blacklist loading of unstable roms

 hw/misc/vfio.c | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 hw/pci/pci.c   |  7 ++++++-
 2 files changed, 64 insertions(+), 1 deletion(-)
Signature · 1 line
1.8.3.1
author
#1 lore raw

PATCH v2 1/2 pci: change default value of rom_bar to 2

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +6 -1
index 4e0701d..12c3e27 100644
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,
Signature · 1 line
1.8.3.1
#2 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

Quoted 19 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +0 -0
index 4e0701d..12c3e27 100644
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 64

Quoted 10 lines
hw/pci/pci.c +6 -1
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,

A slightly more satisfying option might be to define rom_bar as int32_t with default of -1. I don't know if that would break libvirt though. I'll let MST weigh in. Thanks,

Alex

author
#3 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

Quoted 26 lines · Alex Williamson
Quoted 25 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +5 -1
index 4e0701d..12c3e27 100644
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
Quoted 8 lines

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 64

Quoted 4 lines
hw/pci/pci.c +1 -0
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,

A slightly more satisfying option might be to define rom_bar as int32_t with default of -1. I don't know if that would break libvirt though. I'll let MST weigh in. Thanks,

IMO, since the default is "enabled", having a value of -1 for an enabled option could be a bit confusing to some.

Bandan

Quoted 1 line

Alex

#4 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

In reply to Alex Williamson
Quoted 35 lines · Alex Williamson
Quoted 19 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +0 -0
index 4e0701d..12c3e27 100644
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 64

Quoted 10 lines
hw/pci/pci.c +6 -1
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,

A slightly more satisfying option might be to define rom_bar as int32_t with default of -1. I don't know if that would break libvirt though. I'll let MST weigh in. Thanks,

Quoted 1 line

Alex

I don't see rombar in json schema at all. I think it was designed as an internal flag for compatibility with legacy machine types. As such it's likely not a good interface for users.

Signature · 1 line
MST
#5 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

In reply to Bandan Das
Quoted 16 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +0 -0
index 4e0701d..12c3e27 100644

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 61

Quoted 10 lines
hw/pci/pci.c +6 -1
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),

How do users figure out this interface?
Read code?
Could we add a bit property rombarforce=on/off instead? Seems better.

Maybe we should teach bool type visitors about 0 and 1 being legal values
(call out to int visitor, then check value 0 or 1), then rombar can be changed to bit property too.

Also, this will need QMP support right? IIUC rombar is not exposed in QMP ATM.

Quoted 5 lines
hw/pci/pci.c +0 -0
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,
Signature · 1 line
1.8.3.1
author
#6 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

Quoted 24 lines · Michael S. Tsirkin
Quoted 23 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +3 -1
index 4e0701d..12c3e27 100644
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
Quoted 8 lines

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 61

Quoted 3 lines
hw/pci/pci.c +3 -0
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),

How do users figure out this interface?
Read code?
Could we add a bit property rombarforce=on/off instead? Seems better.

Fine with me, but aren't rombarforce and rombar a little bit similar in thier functions ?

1. rombarforce=on automatically implies rombar=1 (force)
2. rombarforce=off(default), rombar=1 (on but don't force loading if blacklisted)
3. rombarforce=on, rombar=0 (do nothing)
4. rombarforce=off, rombar=0 (again nothing)

So why not just a tristate rombar=on/off/force

Quoted 7 lines

Maybe we should teach bool type visitors about 0 and 1 being legal values
(call out to int visitor, then check value 0 or 1), then rombar can be changed to bit property too.

Also, this will need QMP support right? IIUC rombar is not exposed in QMP ATM.

Will qmp hotplug care about option rom ? If not, atleast the reboot will, so I think this is needed.

Quoted 5 lines
Quoted 5 lines
hw/pci/pci.c +0 -0
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,
Signature · 1 line
1.8.3.1
#7 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

In reply to Michael S. Tsirkin
Quoted 33 lines · Michael S. Tsirkin
Quoted 16 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +0 -0
index 4e0701d..12c3e27 100644

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 61

Quoted 10 lines
hw/pci/pci.c +6 -1
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),

How do users figure out this interface?
Read code?
Could we add a bit property rombarforce=on/off instead? Seems better.

Quoted 7 lines

Maybe we should teach bool type visitors about 0 and 1 being legal values
(call out to int visitor, then check value 0 or 1), then rombar can be changed to bit property too.

Also, this will need QMP support right? IIUC rombar is not exposed in QMP ATM.

rombarforce seems very redundant for a user interface; rombar=1 "expose the ROM BAR of the device", rombarforce=1 "yes, really expose the ROM BAR of the device". Even if force implies rombar, I don't think that's very easy to code in libvirt. I think we really just want to detect unspecified versus specified, which probably means setting the default value to something the user can't, or at least wouldn't, specify. Thanks,

Alex

Quoted 6 lines
Quoted 6 lines
hw/pci/pci.c +0 -0
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,
Signature · 1 line
1.8.3.1
#8 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

Quoted 39 lines · Alex Williamson
Quoted 38 lines · Michael S. Tsirkin
Quoted 16 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +0 -0
index 4e0701d..12c3e27 100644

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 61

Quoted 10 lines
hw/pci/pci.c +6 -1
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),

How do users figure out this interface?
Read code?
Could we add a bit property rombarforce=on/off instead? Seems better.

Maybe we should teach bool type visitors about 0 and 1 being legal values
(call out to int visitor, then check value 0 or 1), then rombar can be changed to bit property too.

Quoted 6 lines
Quoted 2 lines

Also, this will need QMP support right? IIUC rombar is not exposed in QMP ATM.

rombarforce seems very redundant for a user interface; rombar=1 "expose the ROM BAR of the device", rombarforce=1 "yes, really expose the ROM BAR of the device".

Not really.
In this design, rombarforce=yes means "expose ROM BAR of the device", rombar should not be exposed to users - it's a compatibility property used for cross-version migration.

Quoted 3 lines

Even if force implies rombar,
I don't think that's
very easy to code in libvirt.

Libvirt doesn't touch rombar AFAIK.

Quoted 6 lines

I think we really just want to detect
unspecified versus specified, which probably means setting the default value to something the user can't, or at least wouldn't, specify. Thanks,

Alex

OK but I should be able to query value of each variable and figure out what it means.

We can build a tri-state property type if desired:
force on/force off/auto.
Just let's not code up random magic values. 0 and 1 for on/off is ugly enough.

Quoted 8 lines
Quoted 6 lines
Quoted 6 lines
hw/pci/pci.c +0 -0
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,
Signature · 1 line
1.8.3.1
#9 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

Quoted 52 lines · Michael S. Tsirkin
Quoted 39 lines · Alex Williamson
Quoted 38 lines · Michael S. Tsirkin
Quoted 16 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +0 -0
index 4e0701d..12c3e27 100644

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 61

Quoted 10 lines
hw/pci/pci.c +6 -1
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),

How do users figure out this interface?
Read code?
Could we add a bit property rombarforce=on/off instead? Seems better.

Maybe we should teach bool type visitors about 0 and 1 being legal values
(call out to int visitor, then check value 0 or 1), then rombar can be changed to bit property too.

Quoted 6 lines
Quoted 2 lines

Also, this will need QMP support right? IIUC rombar is not exposed in QMP ATM.

rombarforce seems very redundant for a user interface; rombar=1 "expose the ROM BAR of the device", rombarforce=1 "yes, really expose the ROM BAR of the device".

Not really.
In this design, rombarforce=yes means "expose ROM BAR of the device", rombar should not be exposed to users - it's a compatibility property used for cross-version migration.

Quoted 5 lines
Quoted 3 lines

Even if force implies rombar,
I don't think that's
very easy to code in libvirt.

Libvirt doesn't touch rombar AFAIK.

It does

http://libvirt.org/formatdomain.html#elementsNICSROM

<rom bar='off'>

Quoted 24 lines
Quoted 6 lines

I think we really just want to detect
unspecified versus specified, which probably means setting the default value to something the user can't, or at least wouldn't, specify. Thanks,

Alex

OK but I should be able to query value of each variable and figure out what it means.

We can build a tri-state property type if desired:
force on/force off/auto.
Just let's not code up random magic values. 0 and 1 for on/off is ugly enough.

Quoted 8 lines
Quoted 6 lines
Quoted 6 lines
hw/pci/pci.c +0 -0
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,
Signature · 1 line
1.8.3.1
#10 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

Quoted 57 lines · Alex Williamson
Quoted 56 lines · Michael S. Tsirkin
Quoted 39 lines · Alex Williamson
Quoted 38 lines · Michael S. Tsirkin
Quoted 16 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +0 -0
index 4e0701d..12c3e27 100644

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 61

Quoted 10 lines
hw/pci/pci.c +6 -1
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),

How do users figure out this interface?
Read code?
Could we add a bit property rombarforce=on/off instead? Seems better.

Maybe we should teach bool type visitors about 0 and 1 being legal values
(call out to int visitor, then check value 0 or 1), then rombar can be changed to bit property too.

Quoted 6 lines
Quoted 2 lines

Also, this will need QMP support right? IIUC rombar is not exposed in QMP ATM.

rombarforce seems very redundant for a user interface; rombar=1 "expose the ROM BAR of the device", rombarforce=1 "yes, really expose the ROM BAR of the device".

Not really.
In this design, rombarforce=yes means "expose ROM BAR of the device", rombar should not be exposed to users - it's a compatibility property used for cross-version migration.

Quoted 3 lines

Even if force implies rombar,
I don't think that's
very easy to code in libvirt.

Quoted 7 lines
Quoted 1 line

Libvirt doesn't touch rombar AFAIK.

It does

http://libvirt.org/formatdomain.html#elementsNICSROM

<rom bar='off'>

Got it, thanks. So if you think the right thing to do for users it to interpret rom=on as meaning "force" then just do that.
Use some new hidden field for machine compatibility.

Quoted 26 lines
Quoted 24 lines
Quoted 6 lines

I think we really just want to detect
unspecified versus specified, which probably means setting the default value to something the user can't, or at least wouldn't, specify. Thanks,

Alex

OK but I should be able to query value of each variable and figure out what it means.

We can build a tri-state property type if desired:
force on/force off/auto.
Just let's not code up random magic values. 0 and 1 for on/off is ugly enough.

Quoted 8 lines
Quoted 6 lines
Quoted 6 lines
hw/pci/pci.c +0 -0
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,
Signature · 1 line
1.8.3.1
author
#11 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

Quoted 64 lines · Michael S. Tsirkin
Quoted 63 lines · Alex Williamson
Quoted 52 lines · Michael S. Tsirkin
Quoted 39 lines · Alex Williamson
Quoted 38 lines · Michael S. Tsirkin
Quoted 16 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +0 -0
index 4e0701d..12c3e27 100644

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 61

Quoted 10 lines
hw/pci/pci.c +6 -1
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),

How do users figure out this interface?
Read code?
Could we add a bit property rombarforce=on/off instead? Seems better.

Maybe we should teach bool type visitors about 0 and 1 being legal values
(call out to int visitor, then check value 0 or 1), then rombar can be changed to bit property too.

Quoted 6 lines
Quoted 2 lines

Also, this will need QMP support right? IIUC rombar is not exposed in QMP ATM.

rombarforce seems very redundant for a user interface; rombar=1 "expose the ROM BAR of the device", rombarforce=1 "yes, really expose the ROM BAR of the device".

Not really.
In this design, rombarforce=yes means "expose ROM BAR of the device", rombar should not be exposed to users - it's a compatibility property used for cross-version migration.

Quoted 5 lines
Quoted 3 lines

Even if force implies rombar,
I don't think that's
very easy to code in libvirt.

Libvirt doesn't touch rombar AFAIK.

It does

http://libvirt.org/formatdomain.html#elementsNICSROM

Quoted 7 lines
Quoted 1 line

<rom bar='off'>

Got it, thanks. So if you think the right thing to do for users it to interpret rom=on as meaning "force" then just do that.
Use some new hidden field for machine compatibility.

Even if we use another variable for machine compatibility, we can't assume rom=on means force.

"force" is that special case where even if the rom is blacklisted, loading is attempted. (Please see 2/2 v2] vfio: blacklist loading of unstable roms) For now, the usecase is to get around when there is a new rom to test.

A tristate property seems better, with an approach that addresses your concerns about random values that could confuse users.

Bandan

Quoted 27 lines
Quoted 27 lines
Quoted 25 lines
Quoted 7 lines

I think we really just want to detect
unspecified versus specified, which probably means setting the default value to something the user can't, or at least wouldn't, specify. Thanks,

Alex

OK but I should be able to query value of each variable and figure out what it means.

We can build a tri-state property type if desired:
force on/force off/auto.
Just let's not code up random magic values. 0 and 1 for on/off is ugly enough.

Quoted 8 lines
Quoted 6 lines
Quoted 6 lines
hw/pci/pci.c +0 -0
     DEFINE_PROP_BIT("multifunction", PCIDevice, cap_present,
                     QEMU_PCI_CAP_MULTIFUNCTION_BITNR, false),
     DEFINE_PROP_BIT("command_serr_enable", PCIDevice, cap_present,
Signature · 1 line
1.8.3.1
#12 lore raw

Re: [Qemu-devel] [PATCH 1/2 v2] pci: change default value of rom_bar to 2

Quoted 77 lines · Bandan Das
Quoted 64 lines · Michael S. Tsirkin
Quoted 63 lines · Alex Williamson
Quoted 52 lines · Michael S. Tsirkin
Quoted 39 lines · Alex Williamson
Quoted 38 lines · Michael S. Tsirkin
Quoted 16 lines · Bandan Das

The following patch depends on the value of rom_bar to determine rom blacklist behavior. Existing code shouldn't be affected by changing the default value of rom_bar since all relevant decisions only rely on whether rom_bar is zero or non-zero.

---
 hw/pci/pci.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
hw/pci/pci.c +0 -0
index 4e0701d..12c3e27 100644

hw/pci/pci.c static void pci_bus_finalize(Object *obj); · line 61

Quoted 10 lines
hw/pci/pci.c +6 -1
@@ -53,7 +53,12 @@ static void pci_bus_finalize(Object *obj);
 static Property pci_props[] = {
     DEFINE_PROP_PCI_DEVFN("addr", PCIDevice, devfn, -1),
     DEFINE_PROP_STRING("romfile", PCIDevice, romfile),
-    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 1),
+    /*
+     * 0 = disable
+     * 1 = user requested on, force loading even if rom blacklisted
+     * 2 = enabled but disables loading of blacklisted roms (default)
+     */
+    DEFINE_PROP_UINT32("rombar",  PCIDevice, rom_bar, 2),

How do users figure out this interface?
Read code?
Could we add a bit property rombarforce=on/off instead? Seems better.

Maybe we should teach bool type visitors about 0 and 1 being legal values
(call out to int visitor, then check value 0 or 1), then rombar can be changed to bit property too.

Quoted 6 lines
Quoted 2 lines

Also, this will need QMP support right? IIUC rombar is not exposed in QMP ATM.

rombarforce seems very redundant for a user interface; rombar=1 "expose the ROM BAR of the device", rombarforce=1 "yes, really expose the ROM BAR of the device".

Not really.
In this design, rombarforce=yes means "expose ROM BAR of the device", rombar should not be exposed to users - it's a compatibility property used for cross-version migration.

Quoted 5 lines
Quoted 3 lines

Even if force implies rombar,
I don't think that's
very easy to code in libvirt.

Libvirt doesn't touch rombar AFAIK.

It does

http://libvirt.org/formatdomain.html#elementsNICSROM

Quoted 7 lines
Quoted 1 line

<rom bar='off'>

Got it, thanks. So if you think the right thing to do for users it to interpret rom=on as meaning "force" then just do that.
Use some new hidden field for machine compatibility.

Even if we use another variable for machine compatibility, we can't assume rom=on means force.

Quoted 6 lines

"force" is that special case where even if the rom is blacklisted, loading is attempted. (Please see 2/2 v2] vfio: blacklist loading of unstable roms) For now, the usecase is to get around when there is a new rom to test.

A tristate property seems better, with an approach that addresses your concerns about random values that could confuse users.

I suspect there are ways to parse the opts for a given device to find whether rombar was specified so we don't need to create a magic "unset" value. We just need to dig through the obfuscation of the opts code.

author
#13 lore raw

PATCH v2 2/2 vfio: blacklist loading of unstable roms

In reply to Bandan Das

Certain cards such as the Broadcom BCM57810 have rom quirks that exhibit unstable system behavior duing device assignment. In the particular case of 57810, rom execution hangs and if a FLR follows, the device becomes inoperable until a power cycle. This is a simple change to blacklist loading of rom for such cards unless the user specifies a romfile or rombar=1 on the cmd line

---
 hw/misc/vfio.c | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 58 insertions(+)
hw/misc/vfio.c +58 -0
index 8db182f..58f348e 100644
@@ -209,6 +209,16 @@ typedef struct VFIOGroup {
     QLIST_ENTRY(VFIOGroup) container_next;
 } VFIOGroup;
 
+typedef struct VFIORomBlacklistEntry {
+    uint16_t vendor_id;
+    uint16_t device_id;
+} VFIORomBlacklistEntry;
+
+static const VFIORomBlacklistEntry romblacklist[] = {
+    /* Broadcom BCM 57810 */
+    { 0x14e4, 0x168e }
+};
+
 #define MSIX_CAP_LENGTH 12
 
 static QLIST_HEAD(, VFIOContainer)
@@ -1197,6 +1207,26 @@ static const MemoryRegionOps vfio_rom_ops = {
     .endianness = DEVICE_LITTLE_ENDIAN,
 };
 
+static bool vfio_blacklist_opt_rom(VFIODevice *vdev)
+{
+    PCIDevice *pdev = &vdev->pdev;
+    uint16_t vendor_id, device_id;
+    int count = 0;
+
+    vendor_id = pci_get_word(pdev->config + PCI_VENDOR_ID);
+    device_id = pci_get_word(pdev->config + PCI_DEVICE_ID);
+
+    while (count < ARRAY_SIZE(romblacklist)) {
+        if (romblacklist[count].vendor_id == vendor_id &&
+            romblacklist[count].device_id == device_id) {
+                return true;
+        }
+        count++;
+    }
+
+    return false;
+}
+
 static void vfio_pci_size_rom(VFIODevice *vdev)
 {
     uint32_t orig, size = cpu_to_le32((uint32_t)PCI_ROM_ADDRESS_MASK);
@@ -1204,9 +1234,37 @@ static void vfio_pci_size_rom(VFIODevice *vdev)
     char name[32];
 
     if (vdev->pdev.romfile || !vdev->pdev.rom_bar) {
+        /* Since pci handles romfile, just print a message and return */
+        if (vfio_blacklist_opt_rom(vdev) && vdev->pdev.romfile) {
+            error_printf("Warning : Device at %04x:%02x:%02x.%x "
+                         "is known to cause system instability issues during "
+                         "option rom execution. "
+                         "Proceeding anyway since user specified romfile\n",
+                         vdev->host.domain, vdev->host.bus, vdev->host.slot,
+                         vdev->host.function);
+        }
         return;
     }
 
+    if (vfio_blacklist_opt_rom(vdev) && vdev->pdev.rom_bar) {
+        if (vdev->pdev.rom_bar == 1) {
+            error_printf("Warning : Device at %04x:%02x:%02x.%x "
+                         "is known to cause system instability issues during "
+                         "option rom execution. "
+                         "Proceeding anyway since user specified rombar=1\n",
+                         vdev->host.domain, vdev->host.bus, vdev->host.slot,
+                         vdev->host.function);
+        } else {
+            error_printf("Warning : Rom loading for device at "
+                         "%04x:%02x:%02x.%x has been disabled due to "
+                         "system instability issues. "
+                         "Specify rombar=1 or romfile to force\n",
+                         vdev->host.domain, vdev->host.bus, vdev->host.slot,
+                         vdev->host.function);
+            return;
+        }
+    }
+
     /*
      * Use the same size ROM BAR as the physical device.  The contents
      * will get filled in later when the guest tries to read it.
Signature · 1 line
1.8.3.1
#14 lore raw
Quoted 20 lines · Bandan Das

Certain cards such as the Broadcom BCM57810 have rom quirks that exhibit unstable system behavior duing device assignment. In the particular case of 57810, rom execution hangs and if a FLR follows, the device becomes inoperable until a power cycle. This is a simple change to blacklist loading of rom for such cards unless the user specifies a romfile or rombar=1 on the cmd line

---
 hw/misc/vfio.c | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 58 insertions(+)
hw/misc/vfio.c +0 -0
index 8db182f..58f348e 100644
@@ -209,6 +209,16 @@ typedef struct VFIOGroup {
     QLIST_ENTRY(VFIOGroup) container_next;
 } VFIOGroup;
 

hw/misc/vfio.c typedef struct VFIOGroup { · line 221

Quoted 10 lines
hw/misc/vfio.c +10 -0
+typedef struct VFIORomBlacklistEntry {
+    uint16_t vendor_id;
+    uint16_t device_id;
+} VFIORomBlacklistEntry;
+
+static const VFIORomBlacklistEntry romblacklist[] = {
+    /* Broadcom BCM 57810 */
+    { 0x14e4, 0x168e }
+};
+

Ideally we'd want to be able to reference a bugzilla here so we have some reference in the code to track developments. Also, can we capture the ROM version known to cause this problem so if somebody later says that it works we can have something to compare? The PCI firmware spec defines the data structure. Effectively you can dump the ROM from the device, run xxd on it and look for the "PCIR" string that defines the start of the PCI data structure. The 2 bytes at offset 12h (where 0h is the 'P' in "PCIR") have a revision level field.

Quoted 55 lines
hw/misc/vfio.c +38 -0
 #define MSIX_CAP_LENGTH 12
 
 static QLIST_HEAD(, VFIOContainer)
@@ -1197,6 +1207,26 @@ static const MemoryRegionOps vfio_rom_ops = {
     .endianness = DEVICE_LITTLE_ENDIAN,
 };
 
+static bool vfio_blacklist_opt_rom(VFIODevice *vdev)
+{
+    PCIDevice *pdev = &vdev->pdev;
+    uint16_t vendor_id, device_id;
+    int count = 0;
+
+    vendor_id = pci_get_word(pdev->config + PCI_VENDOR_ID);
+    device_id = pci_get_word(pdev->config + PCI_DEVICE_ID);
+
+    while (count < ARRAY_SIZE(romblacklist)) {
+        if (romblacklist[count].vendor_id == vendor_id &&
+            romblacklist[count].device_id == device_id) {
+                return true;
+        }
+        count++;
+    }
+
+    return false;
+}
+
 static void vfio_pci_size_rom(VFIODevice *vdev)
 {
     uint32_t orig, size = cpu_to_le32((uint32_t)PCI_ROM_ADDRESS_MASK);
@@ -1204,9 +1234,37 @@ static void vfio_pci_size_rom(VFIODevice *vdev)
     char name[32];
 
     if (vdev->pdev.romfile || !vdev->pdev.rom_bar) {
+        /* Since pci handles romfile, just print a message and return */
+        if (vfio_blacklist_opt_rom(vdev) && vdev->pdev.romfile) {
+            error_printf("Warning : Device at %04x:%02x:%02x.%x "
+                         "is known to cause system instability issues during "
+                         "option rom execution. "
+                         "Proceeding anyway since user specified romfile\n",
+                         vdev->host.domain, vdev->host.bus, vdev->host.slot,
+                         vdev->host.function);
+        }
         return;
     }
 
+    if (vfio_blacklist_opt_rom(vdev) && vdev->pdev.rom_bar) {
+        if (vdev->pdev.rom_bar == 1) {
+            error_printf("Warning : Device at %04x:%02x:%02x.%x "
+                         "is known to cause system instability issues during "
+                         "option rom execution. "
+                         "Proceeding anyway since user specified rombar=1\n",
+                         vdev->host.domain, vdev->host.bus, vdev->host.slot,
+                         vdev->host.function);
+        } else {

hw/misc/vfio.c static void vfio_pci_size_rom(VFIODevice *vdev) · line 1267

Quoted 10 lines
hw/misc/vfio.c +10 -0
+            error_printf("Warning : Rom loading for device at "
+                         "%04x:%02x:%02x.%x has been disabled due to "
+                         "system instability issues. "
+                         "Specify rombar=1 or romfile to force\n",
+                         vdev->host.domain, vdev->host.bus, vdev->host.slot,
+                         vdev->host.function);
+            return;
+        }
+    }
+

What happens if this card, or some later user of this blacklisting, has SKUs that don't have a ROM? Aren't we going to print this last warning regardless of whether we found anything to load? Just shifting this whole block down below the next couple tests is probably sufficient. Thanks,

Alex

Quoted 3 lines
hw/misc/vfio.c +0 -0
     /*
      * Use the same size ROM BAR as the physical device.  The contents
      * will get filled in later when the guest tries to read it.
author
#15 lore raw
Quoted 104 lines · Alex Williamson
Quoted 20 lines · Bandan Das

Certain cards such as the Broadcom BCM57810 have rom quirks that exhibit unstable system behavior duing device assignment. In the particular case of 57810, rom execution hangs and if a FLR follows, the device becomes inoperable until a power cycle. This is a simple change to blacklist loading of rom for such cards unless the user specifies a romfile or rombar=1 on the cmd line

---
 hw/misc/vfio.c | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 58 insertions(+)
hw/misc/vfio.c +0 -0
index 8db182f..58f348e 100644
@@ -209,6 +209,16 @@ typedef struct VFIOGroup {
     QLIST_ENTRY(VFIOGroup) container_next;
 } VFIOGroup;
 

hw/misc/vfio.c typedef struct VFIOGroup { · line 221

Quoted 10 lines
hw/misc/vfio.c +10 -0
+typedef struct VFIORomBlacklistEntry {
+    uint16_t vendor_id;
+    uint16_t device_id;
+} VFIORomBlacklistEntry;
+
+static const VFIORomBlacklistEntry romblacklist[] = {
+    /* Broadcom BCM 57810 */
+    { 0x14e4, 0x168e }
+};
+

Ideally we'd want to be able to reference a bugzilla here so we have some reference in the code to track developments. Also, can we capture the ROM version known to cause this problem so if somebody later says that it works we can have something to compare? The PCI firmware spec defines the data structure. Effectively you can dump the ROM from the device, run xxd on it and look for the "PCIR" string that defines the start of the PCI data structure. The 2 bytes at offset 12h (where 0h is the 'P' in "PCIR") have a revision level field.

Quoted 63 lines
hw/misc/vfio.c +46 -0
 #define MSIX_CAP_LENGTH 12
 
 static QLIST_HEAD(, VFIOContainer)
@@ -1197,6 +1207,26 @@ static const MemoryRegionOps vfio_rom_ops = {
     .endianness = DEVICE_LITTLE_ENDIAN,
 };
 
+static bool vfio_blacklist_opt_rom(VFIODevice *vdev)
+{
+    PCIDevice *pdev = &vdev->pdev;
+    uint16_t vendor_id, device_id;
+    int count = 0;
+
+    vendor_id = pci_get_word(pdev->config + PCI_VENDOR_ID);
+    device_id = pci_get_word(pdev->config + PCI_DEVICE_ID);
+
+    while (count < ARRAY_SIZE(romblacklist)) {
+        if (romblacklist[count].vendor_id == vendor_id &&
+            romblacklist[count].device_id == device_id) {
+                return true;
+        }
+        count++;
+    }
+
+    return false;
+}
+
 static void vfio_pci_size_rom(VFIODevice *vdev)
 {
     uint32_t orig, size = cpu_to_le32((uint32_t)PCI_ROM_ADDRESS_MASK);
@@ -1204,9 +1234,37 @@ static void vfio_pci_size_rom(VFIODevice *vdev)
     char name[32];
 
     if (vdev->pdev.romfile || !vdev->pdev.rom_bar) {
+        /* Since pci handles romfile, just print a message and return */
+        if (vfio_blacklist_opt_rom(vdev) && vdev->pdev.romfile) {
+            error_printf("Warning : Device at %04x:%02x:%02x.%x "
+                         "is known to cause system instability issues during "
+                         "option rom execution. "
+                         "Proceeding anyway since user specified romfile\n",
+                         vdev->host.domain, vdev->host.bus, vdev->host.slot,
+                         vdev->host.function);
+        }
         return;
     }
 
+    if (vfio_blacklist_opt_rom(vdev) && vdev->pdev.rom_bar) {
+        if (vdev->pdev.rom_bar == 1) {
+            error_printf("Warning : Device at %04x:%02x:%02x.%x "
+                         "is known to cause system instability issues during "
+                         "option rom execution. "
+                         "Proceeding anyway since user specified rombar=1\n",
+                         vdev->host.domain, vdev->host.bus, vdev->host.slot,
+                         vdev->host.function);
+        } else {
+            error_printf("Warning : Rom loading for device at "
+                         "%04x:%02x:%02x.%x has been disabled due to "
+                         "system instability issues. "
+                         "Specify rombar=1 or romfile to force\n",
+                         vdev->host.domain, vdev->host.bus, vdev->host.slot,
+                         vdev->host.function);
+            return;
+        }
Quoted 8 lines

hw/misc/vfio.c static void vfio_pci_size_rom(VFIODevice *vdev) · line 1267

Quoted 2 lines
hw/misc/vfio.c +2 -0
+    }
+

What happens if this card, or some later user of this blacklisting, has SKUs that don't have a ROM? Aren't we going to print this last warning regardless of whether we found anything to load? Just shifting this whole block down below the next couple tests is probably sufficient. Thanks,

Thanks for catching this! I will shift this down for the next version..

Quoted 5 lines

Alex

Quoted 3 lines
hw/misc/vfio.c +0 -0
     /*
      * Use the same size ROM BAR as the physical device.  The contents
      * will get filled in later when the guest tries to read it.