Introduce a new pcibios function pcibios_ignore_alignment_request
which allows the PCI core to defer to platform-specific code to
determine whether or not to ignore alignment requests for PCI resources.
The existing behavior is to simply ignore alignment requests when
PCI_PROBE_ONLY is set. This is behavior is maintained by the
default implementation of pcibios_ignore_alignment_request.
Signed-off-by: Shawn Anastasio <[email protected]>
---
drivers/pci/pci.c | 9 +++++++--
include/linux/pci.h | 1 +
2 files changed, 8 insertions(+), 2 deletions(-)
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 8abc843b1615..8207a09085d1 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -5882,6 +5882,11 @@ resource_size_t __weak pcibios_default_alignment(void)
return 0;
}
+int __weak pcibios_ignore_alignment_request(void)
+{
+ return pci_has_flag(PCI_PROBE_ONLY);
+}
+
#define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
static DEFINE_SPINLOCK(resource_alignment_lock);
@@ -5906,9 +5911,9 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
p = resource_alignment_param;
if (!*p && !align)
goto out;
- if (pci_has_flag(PCI_PROBE_ONLY)) {
+ if (pcibios_ignore_alignment_request()) {
align = 0;
- pr_info_once("PCI: Ignoring requested alignments (PCI_PROBE_ONLY)\n");
+ pr_info_once("PCI: Ignoring requested alignments\n");
goto out;
}
diff --git a/include/linux/pci.h b/include/linux/pci.h
index 4a5a84d7bdd4..47471dcdbaf9 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -1990,6 +1990,7 @@ static inline void pcibios_penalize_isa_irq(int irq, int active) {}
int pcibios_alloc_irq(struct pci_dev *dev);
void pcibios_free_irq(struct pci_dev *dev);
resource_size_t pcibios_default_alignment(void);
+int pcibios_ignore_alignment_request(void);
#ifdef CONFIG_HIBERNATE_CALLBACKS
extern struct dev_pm_ops pcibios_pm_ops;
--
2.20.1
On Tue, May 28, 2019 at 2:03 PM Shawn Anastasio <[email protected]> wrote:
>
> Introduce a new pcibios function pcibios_ignore_alignment_request
> which allows the PCI core to defer to platform-specific code to
> determine whether or not to ignore alignment requests for PCI resources.
>
> The existing behavior is to simply ignore alignment requests when
> PCI_PROBE_ONLY is set. This is behavior is maintained by the
> default implementation of pcibios_ignore_alignment_request.
>
> Signed-off-by: Shawn Anastasio <[email protected]>
> ---
> drivers/pci/pci.c | 9 +++++++--
> include/linux/pci.h | 1 +
> 2 files changed, 8 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 8abc843b1615..8207a09085d1 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -5882,6 +5882,11 @@ resource_size_t __weak pcibios_default_alignment(void)
> return 0;
> }
>
> +int __weak pcibios_ignore_alignment_request(void)
> +{
> + return pci_has_flag(PCI_PROBE_ONLY);
> +}
> +
> #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
> static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
> static DEFINE_SPINLOCK(resource_alignment_lock);
> @@ -5906,9 +5911,9 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
> p = resource_alignment_param;
> if (!*p && !align)
> goto out;
> - if (pci_has_flag(PCI_PROBE_ONLY)) {
> + if (pcibios_ignore_alignment_request()) {
> align = 0;
> - pr_info_once("PCI: Ignoring requested alignments (PCI_PROBE_ONLY)\n");
> + pr_info_once("PCI: Ignoring requested alignments\n");
> goto out;
> }
I think the logic here is questionable to begin with. If the user has
explicitly requested re-aligning a resource via the command line then
we should probably do it even if PCI_PROBE_ONLY is set. When it breaks
they get to keep the pieces.
That said, the real issue here is that PCI_PROBE_ONLY probably
shouldn't be set under qemu/kvm. Under the other hypervisor (PowerVM)
hotplugged devices are configured by firmware before it's passed to
the guest and we need to keep the FW assignments otherwise things
break. QEMU however doesn't do any BAR assignments and relies on that
being handled by the guest. At boot time this is done by SLOF, but
Linux only keeps SLOF around until it's extracted the device-tree.
Once that's done SLOF gets blown away and the kernel needs to do it's
own BAR assignments. I'm guessing there's a hack in there to make it
work today, but it's a little surprising that it works at all...
IIRC Sam Bobroff was looking at hotplug under pseries recently so he
might have something to add. He's sick at the moment, but I'll ask him
to take a look at this once he's back among the living
> diff --git a/include/linux/pci.h b/include/linux/pci.h
> index 4a5a84d7bdd4..47471dcdbaf9 100644
> --- a/include/linux/pci.h
> +++ b/include/linux/pci.h
> @@ -1990,6 +1990,7 @@ static inline void pcibios_penalize_isa_irq(int irq, int active) {}
> int pcibios_alloc_irq(struct pci_dev *dev);
> void pcibios_free_irq(struct pci_dev *dev);
> resource_size_t pcibios_default_alignment(void);
> +int pcibios_ignore_alignment_request(void);
>
> #ifdef CONFIG_HIBERNATE_CALLBACKS
> extern struct dev_pm_ops pcibios_pm_ops;
> --
> 2.20.1
>
On 28/05/2019 15:36, Oliver wrote:
> On Tue, May 28, 2019 at 2:03 PM Shawn Anastasio <[email protected]> wrote:
>>
>> Introduce a new pcibios function pcibios_ignore_alignment_request
>> which allows the PCI core to defer to platform-specific code to
>> determine whether or not to ignore alignment requests for PCI resources.
>>
>> The existing behavior is to simply ignore alignment requests when
>> PCI_PROBE_ONLY is set. This is behavior is maintained by the
>> default implementation of pcibios_ignore_alignment_request.
>>
>> Signed-off-by: Shawn Anastasio <[email protected]>
>> ---
>> drivers/pci/pci.c | 9 +++++++--
>> include/linux/pci.h | 1 +
>> 2 files changed, 8 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>> index 8abc843b1615..8207a09085d1 100644
>> --- a/drivers/pci/pci.c
>> +++ b/drivers/pci/pci.c
>> @@ -5882,6 +5882,11 @@ resource_size_t __weak pcibios_default_alignment(void)
>> return 0;
>> }
>>
>> +int __weak pcibios_ignore_alignment_request(void)
>> +{
>> + return pci_has_flag(PCI_PROBE_ONLY);
>> +}
>> +
>> #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
>> static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
>> static DEFINE_SPINLOCK(resource_alignment_lock);
>> @@ -5906,9 +5911,9 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
>> p = resource_alignment_param;
>> if (!*p && !align)
>> goto out;
>> - if (pci_has_flag(PCI_PROBE_ONLY)) {
>> + if (pcibios_ignore_alignment_request()) {
>> align = 0;
>> - pr_info_once("PCI: Ignoring requested alignments (PCI_PROBE_ONLY)\n");
>> + pr_info_once("PCI: Ignoring requested alignments\n");
>> goto out;
>> }
>
> I think the logic here is questionable to begin with. If the user has
> explicitly requested re-aligning a resource via the command line then
> we should probably do it even if PCI_PROBE_ONLY is set. When it breaks
> they get to keep the pieces.
>
> That said, the real issue here is that PCI_PROBE_ONLY probably
> shouldn't be set under qemu/kvm. Under the other hypervisor (PowerVM)
> hotplugged devices are configured by firmware before it's passed to
> the guest and we need to keep the FW assignments otherwise things
> break. QEMU however doesn't do any BAR assignments and relies on that
> being handled by the guest. At boot time this is done by SLOF, but
> Linux only keeps SLOF around until it's extracted the device-tree.
> Once that's done SLOF gets blown away and the kernel needs to do it's
> own BAR assignments. I'm guessing there's a hack in there to make it
> work today, but it's a little surprising that it works at all...
The hack is to run a modified qemu-aware "/usr/sbin/rtas_errd" in the
guest which receives an event from qemu (RAS_EPOW from
/proc/interrupts), fetches device tree chunks (and as I understand it -
they come with BARs from phyp but without from qemu) and writes "1" to
"/sys/bus/pci/rescan" which calls pci_assign_resource() eventually:
[c000000006e6f960] [c0000000005f62d4] pci_assign_resource+0x44/0x360
[c000000006e6fa10] [c0000000005f8b54]
assign_requested_resources_sorted+0x84/0x110
[c000000006e6fa60] [c0000000005f9540] __assign_resources_sorted+0xd0/0x750
[c000000006e6fb40] [c0000000005fb2e0]
__pci_bus_assign_resources+0x80/0x280
[c000000006e6fc00] [c0000000005fb95c]
pci_assign_unassigned_bus_resources+0xbc/0x100
[c000000006e6fc60] [c0000000005e3d74] pci_rescan_bus+0x34/0x60
[c000000006e6fc90] [c0000000005f1ef4] rescan_store+0x84/0xc0
[c000000006e6fcd0] [c00000000068060c] bus_attr_store+0x3c/0x60
[c000000006e6fcf0] [c00000000037853c] sysfs_kf_write+0x5c/0x80
>
> IIRC Sam Bobroff was looking at hotplug under pseries recently so he
> might have something to add. He's sick at the moment, but I'll ask him
> to take a look at this once he's back among the living
>
>> diff --git a/include/linux/pci.h b/include/linux/pci.h
>> index 4a5a84d7bdd4..47471dcdbaf9 100644
>> --- a/include/linux/pci.h
>> +++ b/include/linux/pci.h
>> @@ -1990,6 +1990,7 @@ static inline void pcibios_penalize_isa_irq(int irq, int active) {}
>> int pcibios_alloc_irq(struct pci_dev *dev);
>> void pcibios_free_irq(struct pci_dev *dev);
>> resource_size_t pcibios_default_alignment(void);
>> +int pcibios_ignore_alignment_request(void);
>>
>> #ifdef CONFIG_HIBERNATE_CALLBACKS
>> extern struct dev_pm_ops pcibios_pm_ops;
>> --
>> 2.20.1
>>
--
Alexey
On Tue, May 28, 2019 at 03:36:34PM +1000, Oliver wrote:
> On Tue, May 28, 2019 at 2:03 PM Shawn Anastasio <[email protected]> wrote:
> >
> > Introduce a new pcibios function pcibios_ignore_alignment_request
> > which allows the PCI core to defer to platform-specific code to
> > determine whether or not to ignore alignment requests for PCI resources.
> >
> > The existing behavior is to simply ignore alignment requests when
> > PCI_PROBE_ONLY is set. This is behavior is maintained by the
> > default implementation of pcibios_ignore_alignment_request.
> >
> > Signed-off-by: Shawn Anastasio <[email protected]>
> > ---
> > drivers/pci/pci.c | 9 +++++++--
> > include/linux/pci.h | 1 +
> > 2 files changed, 8 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > index 8abc843b1615..8207a09085d1 100644
> > --- a/drivers/pci/pci.c
> > +++ b/drivers/pci/pci.c
> > @@ -5882,6 +5882,11 @@ resource_size_t __weak pcibios_default_alignment(void)
> > return 0;
> > }
> >
> > +int __weak pcibios_ignore_alignment_request(void)
> > +{
> > + return pci_has_flag(PCI_PROBE_ONLY);
> > +}
> > +
> > #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
> > static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
> > static DEFINE_SPINLOCK(resource_alignment_lock);
> > @@ -5906,9 +5911,9 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
> > p = resource_alignment_param;
> > if (!*p && !align)
> > goto out;
> > - if (pci_has_flag(PCI_PROBE_ONLY)) {
> > + if (pcibios_ignore_alignment_request()) {
> > align = 0;
> > - pr_info_once("PCI: Ignoring requested alignments (PCI_PROBE_ONLY)\n");
> > + pr_info_once("PCI: Ignoring requested alignments\n");
> > goto out;
> > }
>
> I think the logic here is questionable to begin with. If the user has
> explicitly requested re-aligning a resource via the command line then
> we should probably do it even if PCI_PROBE_ONLY is set. When it breaks
> they get to keep the pieces.
I agree. I don't like PCI_PROBE_ONLY in the first place. It's a
sledgehammer approach that doesn't tell us which resource assignments
need to be preserved or why. I'd rather use IORESOURCE_PCI_FIXED and
set it for the BARs where there's actually some sort of
hypervisor/firmware/OS dependency.
If there's a way to avoid another pciobios_*() weak function, that
would also be better.
Bjorn
On 28/05/2019 17:39, Shawn Anastasio wrote:
>
>
> On 5/28/19 1:27 AM, Alexey Kardashevskiy wrote:
>>
>>
>> On 28/05/2019 15:36, Oliver wrote:
>>> On Tue, May 28, 2019 at 2:03 PM Shawn Anastasio <[email protected]>
>>> wrote:
>>>>
>>>> Introduce a new pcibios function pcibios_ignore_alignment_request
>>>> which allows the PCI core to defer to platform-specific code to
>>>> determine whether or not to ignore alignment requests for PCI
>>>> resources.
>>>>
>>>> The existing behavior is to simply ignore alignment requests when
>>>> PCI_PROBE_ONLY is set. This is behavior is maintained by the
>>>> default implementation of pcibios_ignore_alignment_request.
>>>>
>>>> Signed-off-by: Shawn Anastasio <[email protected]>
>>>> ---
>>>> drivers/pci/pci.c | 9 +++++++--
>>>> include/linux/pci.h | 1 +
>>>> 2 files changed, 8 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>>>> index 8abc843b1615..8207a09085d1 100644
>>>> --- a/drivers/pci/pci.c
>>>> +++ b/drivers/pci/pci.c
>>>> @@ -5882,6 +5882,11 @@ resource_size_t __weak
>>>> pcibios_default_alignment(void)
>>>> return 0;
>>>> }
>>>>
>>>> +int __weak pcibios_ignore_alignment_request(void)
>>>> +{
>>>> + return pci_has_flag(PCI_PROBE_ONLY);
>>>> +}
>>>> +
>>>> #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
>>>> static char
>>>> resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
>>>> static DEFINE_SPINLOCK(resource_alignment_lock);
>>>> @@ -5906,9 +5911,9 @@ static resource_size_t
>>>> pci_specified_resource_alignment(struct pci_dev *dev,
>>>> p = resource_alignment_param;
>>>> if (!*p && !align)
>>>> goto out;
>>>> - if (pci_has_flag(PCI_PROBE_ONLY)) {
>>>> + if (pcibios_ignore_alignment_request()) {
>>>> align = 0;
>>>> - pr_info_once("PCI: Ignoring requested alignments
>>>> (PCI_PROBE_ONLY)\n");
>>>> + pr_info_once("PCI: Ignoring requested alignments\n");
>>>> goto out;
>>>> }
>>>
>>> I think the logic here is questionable to begin with. If the user has
>>> explicitly requested re-aligning a resource via the command line then
>>> we should probably do it even if PCI_PROBE_ONLY is set. When it breaks
>>> they get to keep the pieces.
>>>
>>> That said, the real issue here is that PCI_PROBE_ONLY probably
>>> shouldn't be set under qemu/kvm. Under the other hypervisor (PowerVM)
>>> hotplugged devices are configured by firmware before it's passed to
>>> the guest and we need to keep the FW assignments otherwise things
>>> break. QEMU however doesn't do any BAR assignments and relies on that
>>> being handled by the guest. At boot time this is done by SLOF, but
>>> Linux only keeps SLOF around until it's extracted the device-tree.
>>> Once that's done SLOF gets blown away and the kernel needs to do it's
>>> own BAR assignments. I'm guessing there's a hack in there to make it
>>> work today, but it's a little surprising that it works at all...
>>
>>
>> The hack is to run a modified qemu-aware "/usr/sbin/rtas_errd" in the
>> guest which receives an event from qemu (RAS_EPOW from
>> /proc/interrupts), fetches device tree chunks (and as I understand it -
>> they come with BARs from phyp but without from qemu) and writes "1" to
>> "/sys/bus/pci/rescan" which calls pci_assign_resource() eventually:
>
> Interesting. Does this mean that the PHYP hotplug path doesn't
> call pci_assign_resource?
I'd expect dlpar_add_slot() to be called under phyp and eventually
pci_device_add() which (I think) may or may not trigger later reassignment.
> If so it means the patch may not
> break that platform after all, though it still may not be
> the correct way of doing things.
We should probably stop enforcing the PCI_PROBE_ONLY flag - it seems
that (unless resource_alignment= is used) the pseries guest should just
walk through all allocated resources and leave them unchanged.
>> [c000000006e6f960] [c0000000005f62d4] pci_assign_resource+0x44/0x360
>>
>> [c000000006e6fa10] [c0000000005f8b54]
>> assign_requested_resources_sorted+0x84/0x110
>> [c000000006e6fa60] [c0000000005f9540]
>> __assign_resources_sorted+0xd0/0x750
>> [c000000006e6fb40] [c0000000005fb2e0]
>> __pci_bus_assign_resources+0x80/0x280
>> [c000000006e6fc00] [c0000000005fb95c]
>> pci_assign_unassigned_bus_resources+0xbc/0x100
>> [c000000006e6fc60] [c0000000005e3d74] pci_rescan_bus+0x34/0x60
>>
>> [c000000006e6fc90] [c0000000005f1ef4] rescan_store+0x84/0xc0
>>
>> [c000000006e6fcd0] [c00000000068060c] bus_attr_store+0x3c/0x60
>>
>> [c000000006e6fcf0] [c00000000037853c] sysfs_kf_write+0x5c/0x80
>>
>>
>>
>>
>>
>>>
>>> IIRC Sam Bobroff was looking at hotplug under pseries recently so he
>>> might have something to add. He's sick at the moment, but I'll ask him
>>> to take a look at this once he's back among the living
>>>
>>>> diff --git a/include/linux/pci.h b/include/linux/pci.h
>>>> index 4a5a84d7bdd4..47471dcdbaf9 100644
>>>> --- a/include/linux/pci.h
>>>> +++ b/include/linux/pci.h
>>>> @@ -1990,6 +1990,7 @@ static inline void
>>>> pcibios_penalize_isa_irq(int irq, int active) {}
>>>> int pcibios_alloc_irq(struct pci_dev *dev);
>>>> void pcibios_free_irq(struct pci_dev *dev);
>>>> resource_size_t pcibios_default_alignment(void);
>>>> +int pcibios_ignore_alignment_request(void);
>>>>
>>>> #ifdef CONFIG_HIBERNATE_CALLBACKS
>>>> extern struct dev_pm_ops pcibios_pm_ops;
>>>> --
>>>> 2.20.1
>>>>
>>
--
Alexey
On Tue, May 28, 2019 at 03:36:34PM +1000, Oliver wrote:
> On Tue, May 28, 2019 at 2:03 PM Shawn Anastasio <[email protected]> wrote:
> >
> > Introduce a new pcibios function pcibios_ignore_alignment_request
> > which allows the PCI core to defer to platform-specific code to
> > determine whether or not to ignore alignment requests for PCI resources.
> >
> > The existing behavior is to simply ignore alignment requests when
> > PCI_PROBE_ONLY is set. This is behavior is maintained by the
> > default implementation of pcibios_ignore_alignment_request.
> >
> > Signed-off-by: Shawn Anastasio <[email protected]>
> > ---
> > drivers/pci/pci.c | 9 +++++++--
> > include/linux/pci.h | 1 +
> > 2 files changed, 8 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > index 8abc843b1615..8207a09085d1 100644
> > --- a/drivers/pci/pci.c
> > +++ b/drivers/pci/pci.c
> > @@ -5882,6 +5882,11 @@ resource_size_t __weak pcibios_default_alignment(void)
> > return 0;
> > }
> >
> > +int __weak pcibios_ignore_alignment_request(void)
> > +{
> > + return pci_has_flag(PCI_PROBE_ONLY);
> > +}
> > +
> > #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
> > static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
> > static DEFINE_SPINLOCK(resource_alignment_lock);
> > @@ -5906,9 +5911,9 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
> > p = resource_alignment_param;
> > if (!*p && !align)
> > goto out;
> > - if (pci_has_flag(PCI_PROBE_ONLY)) {
> > + if (pcibios_ignore_alignment_request()) {
> > align = 0;
> > - pr_info_once("PCI: Ignoring requested alignments (PCI_PROBE_ONLY)\n");
> > + pr_info_once("PCI: Ignoring requested alignments\n");
> > goto out;
> > }
>
> I think the logic here is questionable to begin with. If the user has
> explicitly requested re-aligning a resource via the command line then
> we should probably do it even if PCI_PROBE_ONLY is set. When it breaks
> they get to keep the pieces.
>
> That said, the real issue here is that PCI_PROBE_ONLY probably
> shouldn't be set under qemu/kvm. Under the other hypervisor (PowerVM)
> hotplugged devices are configured by firmware before it's passed to
> the guest and we need to keep the FW assignments otherwise things
> break. QEMU however doesn't do any BAR assignments and relies on that
> being handled by the guest. At boot time this is done by SLOF, but
> Linux only keeps SLOF around until it's extracted the device-tree.
> Once that's done SLOF gets blown away and the kernel needs to do it's
> own BAR assignments. I'm guessing there's a hack in there to make it
> work today, but it's a little surprising that it works at all...
>
> IIRC Sam Bobroff was looking at hotplug under pseries recently so he
> might have something to add. He's sick at the moment, but I'll ask him
> to take a look at this once he's back among the living
There seems to be some code already in the kernel that will disable
PCI_PROBE_ONLY based on a device tree property, so I did a quick test
today and it seems to work. Only a trivial tweak is needed in QEMU to
do it (have spapr_dt_chosen() add a node called "linux,pci-probe-only"
with a value of 0), and that would allow us to set it only for QEMU (and
not PowerVM) if that's what we want to do. Is that useful?
(I haven't done any real testing yet but the guest booted up OK.)
> > diff --git a/include/linux/pci.h b/include/linux/pci.h
> > index 4a5a84d7bdd4..47471dcdbaf9 100644
> > --- a/include/linux/pci.h
> > +++ b/include/linux/pci.h
> > @@ -1990,6 +1990,7 @@ static inline void pcibios_penalize_isa_irq(int irq, int active) {}
> > int pcibios_alloc_irq(struct pci_dev *dev);
> > void pcibios_free_irq(struct pci_dev *dev);
> > resource_size_t pcibios_default_alignment(void);
> > +int pcibios_ignore_alignment_request(void);
> >
> > #ifdef CONFIG_HIBERNATE_CALLBACKS
> > extern struct dev_pm_ops pcibios_pm_ops;
> > --
> > 2.20.1
> >
>
On 31/05/2019 08:49, Shawn Anastasio wrote:
> On 5/29/19 10:39 PM, Alexey Kardashevskiy wrote:
>>
>>
>> On 28/05/2019 17:39, Shawn Anastasio wrote:
>>>
>>>
>>> On 5/28/19 1:27 AM, Alexey Kardashevskiy wrote:
>>>>
>>>>
>>>> On 28/05/2019 15:36, Oliver wrote:
>>>>> On Tue, May 28, 2019 at 2:03 PM Shawn Anastasio <[email protected]>
>>>>> wrote:
>>>>>>
>>>>>> Introduce a new pcibios function pcibios_ignore_alignment_request
>>>>>> which allows the PCI core to defer to platform-specific code to
>>>>>> determine whether or not to ignore alignment requests for PCI
>>>>>> resources.
>>>>>>
>>>>>> The existing behavior is to simply ignore alignment requests when
>>>>>> PCI_PROBE_ONLY is set. This is behavior is maintained by the
>>>>>> default implementation of pcibios_ignore_alignment_request.
>>>>>>
>>>>>> Signed-off-by: Shawn Anastasio <[email protected]>
>>>>>> ---
>>>>>> drivers/pci/pci.c | 9 +++++++--
>>>>>> include/linux/pci.h | 1 +
>>>>>> 2 files changed, 8 insertions(+), 2 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>>>>>> index 8abc843b1615..8207a09085d1 100644
>>>>>> --- a/drivers/pci/pci.c
>>>>>> +++ b/drivers/pci/pci.c
>>>>>> @@ -5882,6 +5882,11 @@ resource_size_t __weak
>>>>>> pcibios_default_alignment(void)
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> +int __weak pcibios_ignore_alignment_request(void)
>>>>>> +{
>>>>>> + return pci_has_flag(PCI_PROBE_ONLY);
>>>>>> +}
>>>>>> +
>>>>>> #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
>>>>>> static char
>>>>>> resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
>>>>>> static DEFINE_SPINLOCK(resource_alignment_lock);
>>>>>> @@ -5906,9 +5911,9 @@ static resource_size_t
>>>>>> pci_specified_resource_alignment(struct pci_dev *dev,
>>>>>> p = resource_alignment_param;
>>>>>> if (!*p && !align)
>>>>>> goto out;
>>>>>> - if (pci_has_flag(PCI_PROBE_ONLY)) {
>>>>>> + if (pcibios_ignore_alignment_request()) {
>>>>>> align = 0;
>>>>>> - pr_info_once("PCI: Ignoring requested alignments
>>>>>> (PCI_PROBE_ONLY)\n");
>>>>>> + pr_info_once("PCI: Ignoring requested alignments\n");
>>>>>> goto out;
>>>>>> }
>>>>>
>>>>> I think the logic here is questionable to begin with. If the user has
>>>>> explicitly requested re-aligning a resource via the command line then
>>>>> we should probably do it even if PCI_PROBE_ONLY is set. When it breaks
>>>>> they get to keep the pieces.
>>>>>
>>>>> That said, the real issue here is that PCI_PROBE_ONLY probably
>>>>> shouldn't be set under qemu/kvm. Under the other hypervisor (PowerVM)
>>>>> hotplugged devices are configured by firmware before it's passed to
>>>>> the guest and we need to keep the FW assignments otherwise things
>>>>> break. QEMU however doesn't do any BAR assignments and relies on that
>>>>> being handled by the guest. At boot time this is done by SLOF, but
>>>>> Linux only keeps SLOF around until it's extracted the device-tree.
>>>>> Once that's done SLOF gets blown away and the kernel needs to do it's
>>>>> own BAR assignments. I'm guessing there's a hack in there to make it
>>>>> work today, but it's a little surprising that it works at all...
>>>>
>>>>
>>>> The hack is to run a modified qemu-aware "/usr/sbin/rtas_errd" in the
>>>> guest which receives an event from qemu (RAS_EPOW from
>>>> /proc/interrupts), fetches device tree chunks (and as I understand it -
>>>> they come with BARs from phyp but without from qemu) and writes "1" to
>>>> "/sys/bus/pci/rescan" which calls pci_assign_resource() eventually:
>>>
>>> Interesting. Does this mean that the PHYP hotplug path doesn't
>>> call pci_assign_resource?
>>
>>
>> I'd expect dlpar_add_slot() to be called under phyp and eventually
>> pci_device_add() which (I think) may or may not trigger later
>> reassignment.
>>
>>
>>> If so it means the patch may not
>>> break that platform after all, though it still may not be
>>> the correct way of doing things.
>>
>>
>> We should probably stop enforcing the PCI_PROBE_ONLY flag - it seems
>> that (unless resource_alignment= is used) the pseries guest should just
>> walk through all allocated resources and leave them unchanged.
>
> If we add a pcibios_default_alignment() implementation like was
> suggested earlier, then it will behave as if the user has
> specified resource_alignment= by default and SLOF's assignments
> won't be honored (I think).
I removed pci_add_flags(PCI_PROBE_ONLY) from pSeries_setup_arch and
tried booting with and without pci=resource_alignment= and I can see no
difference - BARs are still aligned to 64K as programmed in SLOF; if I
hack SLOF to align to 4K or 32K - BARs get packed and the guest leaves
them unchanged.
> I guess it boils down to one question - is it important that we
> observe SLOF's initial BAR assignments?
It isn't if it's SLOF but it is if it's phyp. It used to not
allow/support BAR reassignment and even if it does not, I'd rather avoid
touching them.
> If not, the device tree
> modification that Sam found would work fine here. Otherwise,
> we need a way to honor the initial assignments from SLOF while
> still allowing custom alignments for hotplugged devices, either
> by deferring to the platform code like I do here, unsetting
> PCI_PROBE_ONLY in certain cases or by using IORESOURCE_PCI_FIXED
> like Bjorn suggested.
>
>>
>>
>>>> [c000000006e6f960] [c0000000005f62d4] pci_assign_resource+0x44/0x360
>>>>
>>>> [c000000006e6fa10] [c0000000005f8b54]
>>>> assign_requested_resources_sorted+0x84/0x110
>>>> [c000000006e6fa60] [c0000000005f9540]
>>>> __assign_resources_sorted+0xd0/0x750
>>>> [c000000006e6fb40] [c0000000005fb2e0]
>>>> __pci_bus_assign_resources+0x80/0x280
>>>> [c000000006e6fc00] [c0000000005fb95c]
>>>> pci_assign_unassigned_bus_resources+0xbc/0x100
>>>> [c000000006e6fc60] [c0000000005e3d74] pci_rescan_bus+0x34/0x60
>>>>
>>>> [c000000006e6fc90] [c0000000005f1ef4] rescan_store+0x84/0xc0
>>>>
>>>> [c000000006e6fcd0] [c00000000068060c] bus_attr_store+0x3c/0x60
>>>>
>>>> [c000000006e6fcf0] [c00000000037853c] sysfs_kf_write+0x5c/0x80
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>>
>>>>> IIRC Sam Bobroff was looking at hotplug under pseries recently so he
>>>>> might have something to add. He's sick at the moment, but I'll ask him
>>>>> to take a look at this once he's back among the living
>>>>>
>>>>>> diff --git a/include/linux/pci.h b/include/linux/pci.h
>>>>>> index 4a5a84d7bdd4..47471dcdbaf9 100644
>>>>>> --- a/include/linux/pci.h
>>>>>> +++ b/include/linux/pci.h
>>>>>> @@ -1990,6 +1990,7 @@ static inline void
>>>>>> pcibios_penalize_isa_irq(int irq, int active) {}
>>>>>> int pcibios_alloc_irq(struct pci_dev *dev);
>>>>>> void pcibios_free_irq(struct pci_dev *dev);
>>>>>> resource_size_t pcibios_default_alignment(void);
>>>>>> +int pcibios_ignore_alignment_request(void);
>>>>>>
>>>>>> #ifdef CONFIG_HIBERNATE_CALLBACKS
>>>>>> extern struct dev_pm_ops pcibios_pm_ops;
>>>>>> --
>>>>>> 2.20.1
>>>>>>
>>>>
>>
--
Alexey
On 03/06/2019 12:23, Shawn Anastasio wrote:
>
>
> On 5/30/19 10:56 PM, Alexey Kardashevskiy wrote:
>>
>>
>> On 31/05/2019 08:49, Shawn Anastasio wrote:
>>> On 5/29/19 10:39 PM, Alexey Kardashevskiy wrote:
>>>>
>>>>
>>>> On 28/05/2019 17:39, Shawn Anastasio wrote:
>>>>>
>>>>>
>>>>> On 5/28/19 1:27 AM, Alexey Kardashevskiy wrote:
>>>>>>
>>>>>>
>>>>>> On 28/05/2019 15:36, Oliver wrote:
>>>>>>> On Tue, May 28, 2019 at 2:03 PM Shawn Anastasio <[email protected]>
>>>>>>> wrote:
>>>>>>>>
>>>>>>>> Introduce a new pcibios function pcibios_ignore_alignment_request
>>>>>>>> which allows the PCI core to defer to platform-specific code to
>>>>>>>> determine whether or not to ignore alignment requests for PCI
>>>>>>>> resources.
>>>>>>>>
>>>>>>>> The existing behavior is to simply ignore alignment requests when
>>>>>>>> PCI_PROBE_ONLY is set. This is behavior is maintained by the
>>>>>>>> default implementation of pcibios_ignore_alignment_request.
>>>>>>>>
>>>>>>>> Signed-off-by: Shawn Anastasio <[email protected]>
>>>>>>>> ---
>>>>>>>> drivers/pci/pci.c | 9 +++++++--
>>>>>>>> include/linux/pci.h | 1 +
>>>>>>>> 2 files changed, 8 insertions(+), 2 deletions(-)
>>>>>>>>
>>>>>>>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>>>>>>>> index 8abc843b1615..8207a09085d1 100644
>>>>>>>> --- a/drivers/pci/pci.c
>>>>>>>> +++ b/drivers/pci/pci.c
>>>>>>>> @@ -5882,6 +5882,11 @@ resource_size_t __weak
>>>>>>>> pcibios_default_alignment(void)
>>>>>>>> return 0;
>>>>>>>> }
>>>>>>>>
>>>>>>>> +int __weak pcibios_ignore_alignment_request(void)
>>>>>>>> +{
>>>>>>>> + return pci_has_flag(PCI_PROBE_ONLY);
>>>>>>>> +}
>>>>>>>> +
>>>>>>>> #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
>>>>>>>> static char
>>>>>>>> resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
>>>>>>>> static DEFINE_SPINLOCK(resource_alignment_lock);
>>>>>>>> @@ -5906,9 +5911,9 @@ static resource_size_t
>>>>>>>> pci_specified_resource_alignment(struct pci_dev *dev,
>>>>>>>> p = resource_alignment_param;
>>>>>>>> if (!*p && !align)
>>>>>>>> goto out;
>>>>>>>> - if (pci_has_flag(PCI_PROBE_ONLY)) {
>>>>>>>> + if (pcibios_ignore_alignment_request()) {
>>>>>>>> align = 0;
>>>>>>>> - pr_info_once("PCI: Ignoring requested alignments
>>>>>>>> (PCI_PROBE_ONLY)\n");
>>>>>>>> + pr_info_once("PCI: Ignoring requested
>>>>>>>> alignments\n");
>>>>>>>> goto out;
>>>>>>>> }
>>>>>>>
>>>>>>> I think the logic here is questionable to begin with. If the user
>>>>>>> has
>>>>>>> explicitly requested re-aligning a resource via the command line
>>>>>>> then
>>>>>>> we should probably do it even if PCI_PROBE_ONLY is set. When it
>>>>>>> breaks
>>>>>>> they get to keep the pieces.
>>>>>>>
>>>>>>> That said, the real issue here is that PCI_PROBE_ONLY probably
>>>>>>> shouldn't be set under qemu/kvm. Under the other hypervisor
>>>>>>> (PowerVM)
>>>>>>> hotplugged devices are configured by firmware before it's passed to
>>>>>>> the guest and we need to keep the FW assignments otherwise things
>>>>>>> break. QEMU however doesn't do any BAR assignments and relies on
>>>>>>> that
>>>>>>> being handled by the guest. At boot time this is done by SLOF, but
>>>>>>> Linux only keeps SLOF around until it's extracted the device-tree.
>>>>>>> Once that's done SLOF gets blown away and the kernel needs to do
>>>>>>> it's
>>>>>>> own BAR assignments. I'm guessing there's a hack in there to make it
>>>>>>> work today, but it's a little surprising that it works at all...
>>>>>>
>>>>>>
>>>>>> The hack is to run a modified qemu-aware "/usr/sbin/rtas_errd" in the
>>>>>> guest which receives an event from qemu (RAS_EPOW from
>>>>>> /proc/interrupts), fetches device tree chunks (and as I understand
>>>>>> it -
>>>>>> they come with BARs from phyp but without from qemu) and writes
>>>>>> "1" to
>>>>>> "/sys/bus/pci/rescan" which calls pci_assign_resource() eventually:
>>>>>
>>>>> Interesting. Does this mean that the PHYP hotplug path doesn't
>>>>> call pci_assign_resource?
>>>>
>>>>
>>>> I'd expect dlpar_add_slot() to be called under phyp and eventually
>>>> pci_device_add() which (I think) may or may not trigger later
>>>> reassignment.
>>>>
>>>>
>>>>> If so it means the patch may not
>>>>> break that platform after all, though it still may not be
>>>>> the correct way of doing things.
>>>>
>>>>
>>>> We should probably stop enforcing the PCI_PROBE_ONLY flag - it seems
>>>> that (unless resource_alignment= is used) the pseries guest should just
>>>> walk through all allocated resources and leave them unchanged.
>>>
>>> If we add a pcibios_default_alignment() implementation like was
>>> suggested earlier, then it will behave as if the user has
>>> specified resource_alignment= by default and SLOF's assignments
>>> won't be honored (I think).
>>
>>
>> I removed pci_add_flags(PCI_PROBE_ONLY) from pSeries_setup_arch and
>> tried booting with and without pci=resource_alignment= and I can see no
>> difference - BARs are still aligned to 64K as programmed in SLOF; if I
>> hack SLOF to align to 4K or 32K - BARs get packed and the guest leaves
>> them unchanged.
>>
>>
>>> I guess it boils down to one question - is it important that we
>>> observe SLOF's initial BAR assignments?
>>
>> It isn't if it's SLOF but it is if it's phyp. It used to not
>> allow/support BAR reassignment and even if it does not, I'd rather avoid
>> touching them.
>
> A quick update. I tried removing pci_add_flags(PCI_PROBE_ONLY) which
> worked, but if I add an implementation of pcibios_default_alignment
> which simply returns PAGE_SIZE, my VM fails to boot and many errors
> from the virtio disk driver are printed to the console.
>
> After some investigation, it seems that with pcibios_default_alignment
> present, Linux will reallocate all resources provided by SLOF on
> boot. I'm still not sure why exactly this causes the virtio driver
> to fail, but it does indicate that there is a reason to keep
> SLOF's initial assignments.
>
> Anybody have an idea what's causing this?
With your changes the guest feels the urge to reassign bars (no idea why
but ok), when it does so, it puts both BARs (one is prefetchable) into
the 32bit non-prefetchable window of the PHB (SLOF puts the prefetchable
bar to a 64bit prefetchable window, I have no idea why the guest does it
different either but this must still work) and then qemu does not
emulate something properly - unassigned_mem_accepts() is triggered on
the bar access - no idea why - I am debugging it right now.
--
Alexey
On 03/06/2019 15:02, Alexey Kardashevskiy wrote:
>
>
> On 03/06/2019 12:23, Shawn Anastasio wrote:
>>
>>
>> On 5/30/19 10:56 PM, Alexey Kardashevskiy wrote:
>>>
>>>
>>> On 31/05/2019 08:49, Shawn Anastasio wrote:
>>>> On 5/29/19 10:39 PM, Alexey Kardashevskiy wrote:
>>>>>
>>>>>
>>>>> On 28/05/2019 17:39, Shawn Anastasio wrote:
>>>>>>
>>>>>>
>>>>>> On 5/28/19 1:27 AM, Alexey Kardashevskiy wrote:
>>>>>>>
>>>>>>>
>>>>>>> On 28/05/2019 15:36, Oliver wrote:
>>>>>>>> On Tue, May 28, 2019 at 2:03 PM Shawn Anastasio <[email protected]>
>>>>>>>> wrote:
>>>>>>>>>
>>>>>>>>> Introduce a new pcibios function pcibios_ignore_alignment_request
>>>>>>>>> which allows the PCI core to defer to platform-specific code to
>>>>>>>>> determine whether or not to ignore alignment requests for PCI
>>>>>>>>> resources.
>>>>>>>>>
>>>>>>>>> The existing behavior is to simply ignore alignment requests when
>>>>>>>>> PCI_PROBE_ONLY is set. This is behavior is maintained by the
>>>>>>>>> default implementation of pcibios_ignore_alignment_request.
>>>>>>>>>
>>>>>>>>> Signed-off-by: Shawn Anastasio <[email protected]>
>>>>>>>>> ---
>>>>>>>>> drivers/pci/pci.c | 9 +++++++--
>>>>>>>>> include/linux/pci.h | 1 +
>>>>>>>>> 2 files changed, 8 insertions(+), 2 deletions(-)
>>>>>>>>>
>>>>>>>>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>>>>>>>>> index 8abc843b1615..8207a09085d1 100644
>>>>>>>>> --- a/drivers/pci/pci.c
>>>>>>>>> +++ b/drivers/pci/pci.c
>>>>>>>>> @@ -5882,6 +5882,11 @@ resource_size_t __weak
>>>>>>>>> pcibios_default_alignment(void)
>>>>>>>>> return 0;
>>>>>>>>> }
>>>>>>>>>
>>>>>>>>> +int __weak pcibios_ignore_alignment_request(void)
>>>>>>>>> +{
>>>>>>>>> + return pci_has_flag(PCI_PROBE_ONLY);
>>>>>>>>> +}
>>>>>>>>> +
>>>>>>>>> #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
>>>>>>>>> static char
>>>>>>>>> resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
>>>>>>>>> static DEFINE_SPINLOCK(resource_alignment_lock);
>>>>>>>>> @@ -5906,9 +5911,9 @@ static resource_size_t
>>>>>>>>> pci_specified_resource_alignment(struct pci_dev *dev,
>>>>>>>>> p = resource_alignment_param;
>>>>>>>>> if (!*p && !align)
>>>>>>>>> goto out;
>>>>>>>>> - if (pci_has_flag(PCI_PROBE_ONLY)) {
>>>>>>>>> + if (pcibios_ignore_alignment_request()) {
>>>>>>>>> align = 0;
>>>>>>>>> - pr_info_once("PCI: Ignoring requested alignments
>>>>>>>>> (PCI_PROBE_ONLY)\n");
>>>>>>>>> + pr_info_once("PCI: Ignoring requested
>>>>>>>>> alignments\n");
>>>>>>>>> goto out;
>>>>>>>>> }
>>>>>>>>
>>>>>>>> I think the logic here is questionable to begin with. If the user
>>>>>>>> has
>>>>>>>> explicitly requested re-aligning a resource via the command line
>>>>>>>> then
>>>>>>>> we should probably do it even if PCI_PROBE_ONLY is set. When it
>>>>>>>> breaks
>>>>>>>> they get to keep the pieces.
>>>>>>>>
>>>>>>>> That said, the real issue here is that PCI_PROBE_ONLY probably
>>>>>>>> shouldn't be set under qemu/kvm. Under the other hypervisor
>>>>>>>> (PowerVM)
>>>>>>>> hotplugged devices are configured by firmware before it's passed to
>>>>>>>> the guest and we need to keep the FW assignments otherwise things
>>>>>>>> break. QEMU however doesn't do any BAR assignments and relies on
>>>>>>>> that
>>>>>>>> being handled by the guest. At boot time this is done by SLOF, but
>>>>>>>> Linux only keeps SLOF around until it's extracted the device-tree.
>>>>>>>> Once that's done SLOF gets blown away and the kernel needs to do
>>>>>>>> it's
>>>>>>>> own BAR assignments. I'm guessing there's a hack in there to make it
>>>>>>>> work today, but it's a little surprising that it works at all...
>>>>>>>
>>>>>>>
>>>>>>> The hack is to run a modified qemu-aware "/usr/sbin/rtas_errd" in the
>>>>>>> guest which receives an event from qemu (RAS_EPOW from
>>>>>>> /proc/interrupts), fetches device tree chunks (and as I understand
>>>>>>> it -
>>>>>>> they come with BARs from phyp but without from qemu) and writes
>>>>>>> "1" to
>>>>>>> "/sys/bus/pci/rescan" which calls pci_assign_resource() eventually:
>>>>>>
>>>>>> Interesting. Does this mean that the PHYP hotplug path doesn't
>>>>>> call pci_assign_resource?
>>>>>
>>>>>
>>>>> I'd expect dlpar_add_slot() to be called under phyp and eventually
>>>>> pci_device_add() which (I think) may or may not trigger later
>>>>> reassignment.
>>>>>
>>>>>
>>>>>> If so it means the patch may not
>>>>>> break that platform after all, though it still may not be
>>>>>> the correct way of doing things.
>>>>>
>>>>>
>>>>> We should probably stop enforcing the PCI_PROBE_ONLY flag - it seems
>>>>> that (unless resource_alignment= is used) the pseries guest should just
>>>>> walk through all allocated resources and leave them unchanged.
>>>>
>>>> If we add a pcibios_default_alignment() implementation like was
>>>> suggested earlier, then it will behave as if the user has
>>>> specified resource_alignment= by default and SLOF's assignments
>>>> won't be honored (I think).
>>>
>>>
>>> I removed pci_add_flags(PCI_PROBE_ONLY) from pSeries_setup_arch and
>>> tried booting with and without pci=resource_alignment= and I can see no
>>> difference - BARs are still aligned to 64K as programmed in SLOF; if I
>>> hack SLOF to align to 4K or 32K - BARs get packed and the guest leaves
>>> them unchanged.
>>>
>>>
>>>> I guess it boils down to one question - is it important that we
>>>> observe SLOF's initial BAR assignments?
>>>
>>> It isn't if it's SLOF but it is if it's phyp. It used to not
>>> allow/support BAR reassignment and even if it does not, I'd rather avoid
>>> touching them.
>>
>> A quick update. I tried removing pci_add_flags(PCI_PROBE_ONLY) which
>> worked, but if I add an implementation of pcibios_default_alignment
>> which simply returns PAGE_SIZE, my VM fails to boot and many errors
>> from the virtio disk driver are printed to the console.
>>
>> After some investigation, it seems that with pcibios_default_alignment
>> present, Linux will reallocate all resources provided by SLOF on
>> boot. I'm still not sure why exactly this causes the virtio driver
>> to fail, but it does indicate that there is a reason to keep
>> SLOF's initial assignments.
>>
>> Anybody have an idea what's causing this?
>
> With your changes the guest feels the urge to reassign bars (no idea why
> but ok), when it does so, it puts both BARs (one is prefetchable) into
> the 32bit non-prefetchable window of the PHB (SLOF puts the prefetchable
> bar to a 64bit prefetchable window, I have no idea why the guest does it
> different either but this must still work) and then qemu does not
> emulate something properly - unassigned_mem_accepts() is triggered on
> the bar access - no idea why - I am debugging it right now.
Sooo the problem is that resouce::flags has 2 bits to describe 64bit
BARs - PCI_BASE_ADDRESS_MEM_TYPE_64 and IORESOURCE_MEM_64 - and we don't
set IORESOURCE_MEM_64 for 64bit BARs when parsing the fdt.
So the BAR reallocator moves the BAR to 32bit window (which is not
desirable but permitted, I still have to chase it) and then
pci_std_update_resource() writes BAR back but since now it is 32bit BAR,
it does not write to the upper 32bits so that half remains 0x2100, QEMU
does not move BAR to the right window and the MMIO stops working.
Try this in the guest kernel, it seems to keep bars where they were
after slof.
diff --git a/arch/powerpc/kernel/pci_of_scan.c
b/arch/powerpc/kernel/pci_of_scan.c
index 24191ea2d9a7..64ad92016b63 100644
--- a/arch/powerpc/kernel/pci_of_scan.c
+++ b/arch/powerpc/kernel/pci_of_scan.c
@@ -45,6 +45,8 @@ unsigned int pci_parse_of_flags(u32 addr0, int bridge)
if (addr0 & 0x02000000) {
flags = IORESOURCE_MEM | PCI_BASE_ADDRESS_SPACE_MEMORY;
flags |= (addr0 >> 22) & PCI_BASE_ADDRESS_MEM_TYPE_64;
+ if (flags & PCI_BASE_ADDRESS_MEM_TYPE_64)
+ flags |= IORESOURCE_MEM_64;
flags |= (addr0 >> 28) & PCI_BASE_ADDRESS_MEM_TYPE_1M;
if (addr0 & 0x40000000)
--
Alexey