Re: [RFC PATCH v3 1/5] PCI: Add support for enforcing all MMIO BARs to be page aligned

[Alex Williamson]

On Fri, 2016-01-29 at 18:37 +0800, Yongji Xie wrote:
> On 2016/1/29 6:46, Alex Williamson wrote:
> > On Fri, 2016-01-15 at 15:06 +0800, Yongji Xie wrote:
> > > When vfio passthrough a PCI device of which MMIO BARs
> > > are smaller than PAGE_SIZE, guest will not handle the
> > > mmio accesses to the BARs which leads to mmio emulations
> > > in host.
> > > ÂÂ
> > > This is because vfio will not allow to passthrough one
> > > BAR's mmio page which may be shared with other BARs.
> > > ÂÂ
> > > To solve this performance issue, this patch adds a kernel
> > > parameter "pci=resource_page_aligned=on" to enforce
> > > the alignment of all MMIO BARs to be at least PAGE_SIZE,
> > > so that one BAR's mmio page would not be shared with other
> > > BARs. We can also disable it through kernel parameter
> > > "pci=resource_page_aligned=off".
> > > ÂÂ
> > > For the default value of the parameter, we think it should be
> > > arch-independent, so we add a macro
> > > HAVE_PCI_DEFAULT_RESOURCES_PAGE_ALIGNED to change it. And we
> > > define this macro to enable this parameter by default on PPC64
> > > platform which can easily hit this performance issue because
> > > its PAGE_SIZE is 64KB.
> > > ÂÂ
> > > Note that the kernel parameter won't works if kernel doesn't do
> > > resources reallocation.
> > And where do you account for this so that we know whether it's really in
> > effect?
>Â
> We can check the flag PCI_PROBE_ONLY to know whether kernel do
> resources reallocation. Then we know if the kernel parameter is really
> in effect.
>Â
> enum {
> ÂÂÂÂÂ/* Force re-assigning all resources (ignore firmware
> ÂÂÂÂÂÂ* setup completely)
> ÂÂÂÂÂÂ*/
> ÂÂÂÂÂPCI_REASSIGN_ALL_RSRCÂÂÂÂ= 0x00000001,
>Â
> ÂÂÂÂÂ/* Re-assign all bus numbers */
> ÂÂÂÂÂPCI_REASSIGN_ALL_BUSÂÂÂÂ= 0x00000002,
>Â
> ÂÂÂÂÂ/* Do not try to assign, just use existing setup */
> --->ÂÂÂÂPCI_PROBE_ONLYÂÂÂÂÂÂÂÂ= 0x00000004,
>Â
> And I will add this to commit log.

We need more than a commit log entry for this, what's the purpose of the
pci_resources_share_page() function if we don't know if this is in
effect?

> > > Signed-off-by: Yongji Xie <xyjxie@xxxxxxxxxxxxxxxxxx>
> > > ---
> > > Â Documentation/kernel-parameters.txt |ÂÂÂÂ5 +++++
> > > Â arch/powerpc/include/asm/pci.hÂÂÂÂÂÂ|ÂÂÂ11 +++++++++++
> > > Â drivers/pci/pci.cÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ|ÂÂÂ35 +++++++++++++++++++++++++++++++++++
> > > Â drivers/pci/pci.hÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ|ÂÂÂÂ8 +++++++-
> > > Â include/linux/pci.hÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ|ÂÂÂÂ4 ++++
> > > Â 5 files changed, 62 insertions(+), 1 deletion(-)
> > > ÂÂ
> > > diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
> > > index 742f69d..3f2a7c9 100644
> > > --- a/Documentation/kernel-parameters.txt
> > > +++ b/Documentation/kernel-parameters.txt
> > > @@ -2857,6 +2857,11 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
> > > ÂÂ				PAGE_SIZE is used as alignment.
> > > ÂÂ				PCI-PCI bridge can be specified, if resource
> > > ÂÂ				windows need to be expanded.
> > > +		resource_page_aligned=	Enable/disable enforcing the alignment
> > > +				of all PCI devices' memory resources to be
> > > +				at least PAGE_SIZE if resources reallocation
> > > +				is done by kernel.
> > > +				Format: { "on" | "off" }
> > > ÂÂ		ecrc=		Enable/disable PCIe ECRC (transaction layer
> > > ÂÂ				end-to-end CRC checking).
> > > ÂÂ				bios: Use BIOS/firmware settings. This is the
> > > diff --git a/arch/powerpc/include/asm/pci.h b/arch/powerpc/include/asm/pci.h
> > > index 3453bd8..2d2b3ef 100644
> > > --- a/arch/powerpc/include/asm/pci.h
> > > +++ b/arch/powerpc/include/asm/pci.h
> > > @@ -136,6 +136,17 @@ extern pgprot_t	pci_phys_mem_access_prot(struct file *file,
> > > ÂÂ					Âunsigned long pfn,
> > > ÂÂ					Âunsigned long size,
> > > ÂÂ					Âpgprot_t prot);
> > > +#ifdef CONFIG_PPC64
> > > +
> > > +/* For PPC64, We enforce all PCI MMIO BARs to be page aligned
> > > + * by default. This would be helpful to improve performance
> > > + * when we passthrough a PCI device of which BARs are smaller
> > > + * than PAGE_SIZE(64KB). And we can use kernel parameter
> > > + * "pci=resource_page_aligned=off" to disable it.
> > > + */
> > > +#define HAVE_PCI_DEFAULT_RESOURCES_PAGE_ALIGNED	1
> > > +
> > > +#endif
> > > ÂÂ
> > > Â #define HAVE_ARCH_PCI_RESOURCE_TO_USER
> > > Â extern void pci_resource_to_user(const struct pci_dev *dev, int bar,
> > > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > > index 314db8c..7b21238 100644
> > > --- a/drivers/pci/pci.c
> > > +++ b/drivers/pci/pci.c
> > > @@ -99,6 +99,9 @@ u8 pci_cache_line_size;
> > > ÂÂÂ*/
> > > Â unsigned int pcibios_max_latency = 255;
> > > ÂÂ
> > > +bool pci_resources_page_aligned =
> > > +	IS_ENABLED(HAVE_PCI_DEFAULT_RESOURCES_PAGE_ALIGNED);
> > I don't think this is proper use of IS_ENABLED, which seems to be
> > targeted at CONFIG_ type options.ÂÂYou could define this as that in an
> > arch Kconfig.
>Â
> Is it better that we define this as a pci Kconfig and select it in archÂ
> Kconfig?

If you want to use IS_ENABLE here, I would think so.

> > > +
> > > Â /* If set, the PCIe ARI capability will not be used. */
> > > Â static bool pcie_ari_disabled;
> > > ÂÂ
> > > @@ -4746,6 +4749,35 @@ static ssize_t pci_resource_alignment_store(struct bus_type *bus,
> > > Â BUS_ATTR(resource_alignment, 0644, pci_resource_alignment_show,
> > > ÂÂ					pci_resource_alignment_store);
> > > ÂÂ
> > > +static void pci_resources_get_page_aligned(char *str)
> > > +{
> > > +	if (!strncmp(str, "off", 3))
> > > +		pci_resources_page_aligned = false;
> > > +	else if (!strncmp(str, "on", 2))
> > > +		pci_resources_page_aligned = true;
> > > +}
> > "get"?
>Â
> How about pci_resources_parse_page_aligned_param()?

Better.

> > > +
> > > +/*
> > > + * This function checks whether PCI BARs' mmio page will be shared
> > > + * with other BARs.
> > > + */
> > > +bool pci_resources_share_page(struct pci_dev *dev, int resno)
> > > +{
> > > +	struct resource *res = dev->resource + resno;
> > > +
> > > +	if (resource_size(res) >= PAGE_SIZE)
> > > +		return false;
> > > +	if (pci_resources_page_aligned && !(res->start & ~PAGE_MASK) &&
> > > +	ÂÂÂÂres->flags & IORESOURCE_MEM) {
> > > +		if (res->sibling)
> > > +			return (res->sibling->start & ~PAGE_MASK);
> > > +		else
> > > +			return false;
> > > +	}
> > > +	return true;
> > > +}
> > > +EXPORT_SYMBOL_GPL(pci_resources_share_page);
> > > +
> > > Â static int __init pci_resource_alignment_sysfs_init(void)
> > > Â {
> > > ÂÂ	return bus_create_file(&pci_bus_type,
> > > @@ -4859,6 +4891,9 @@ static int __init pci_setup(char *str)
> > > ÂÂ			} else if (!strncmp(str, "resource_alignment=", 19)) {
> > > ÂÂ				pci_set_resource_alignment_param(str + 19,
> > > ÂÂ							strlen(str + 19));
> > > +			} else if (!strncmp(str, "resource_page_aligned=",
> > > +				ÂÂÂ22)) {
> > > +				pci_resources_get_page_aligned(str + 22);
> > Doesn't this seem rather redundant with the option right above it,
> > resource_alignment=?ÂÂWhy not modify that to support syntax where all
> > devices get the same alignment?
> >Â
>Â
> The kernel option will be used to do two things.
>Â
> Firstly, the option can be used to enable all devices to be page aligned.
>Â
> Secondly, we can use the option to disable it when the Kconfig option
> mentioned above enable all devices to be page aligned by default.
>Â
> We can easily modify this option "resource_alignment=" to do the first
> thing. But I didn't find a proper way to modify it to do the second thing.

You could allow an arch specified default which is overridden by
specifying a resource_alignment= value.ÂÂThen you need a way to disable
it, which you could simply do by making
pci_set_resource_alignment_param() able to parse something like
resource_alignment=off.

> > > ÂÂ			} else if (!strncmp(str, "ecrc=", 5)) {
> > > ÂÂ				pcie_ecrc_get_policy(str + 5);
> > > ÂÂ			} else if (!strncmp(str, "hpiosize=", 9)) {
> > > diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
> > > index d390fc1..b9b333d 100644
> > > --- a/drivers/pci/pci.h
> > > +++ b/drivers/pci/pci.h
> > > @@ -312,11 +312,17 @@ static inline resource_size_t pci_resource_alignment(struct pci_dev *dev,
> > > Â #ifdef CONFIG_PCI_IOV
> > > ÂÂ	int resno = res - dev->resource;
> > > ÂÂ
> > > -	if (resno >= PCI_IOV_RESOURCES && resno <= PCI_IOV_RESOURCE_END)
> > > +	if (resno >= PCI_IOV_RESOURCES && resno <= PCI_IOV_RESOURCE_END) {
> > > +		if (pci_resources_page_aligned && res->flags & IORESOURCE_MEM)
> > > +			return PAGE_ALIGN(pci_sriov_resource_alignment(dev,
> > > +					ÂÂresno));
> > > ÂÂ		return pci_sriov_resource_alignment(dev, resno);
> > > +	}
> > > Â #endif
> > > ÂÂ	if (dev->class >> 8ÂÂ== PCI_CLASS_BRIDGE_CARDBUS)
> > > ÂÂ		return pci_cardbus_resource_alignment(res);
> > > +	if (pci_resources_page_aligned && res->flags & IORESOURCE_MEM)
> > > +		return PAGE_ALIGN(resource_alignment(res));
> > > ÂÂ	return resource_alignment(res);
> > > Â }
> >Â
> > Since we already have resource_alignment=, shouldn't we already have the
> > code in place to re-align?
>Â
> Yes, this code can do the re-aligning. But we can't reuse the code because
> it re-align device's bars by changing their sizes, which can potentiallyÂ
> break
> some drivers.
>Â
> I'm thinking if we can use IORESOURCE_STARTALIGN for this. Thanks.

Shouldn't we fix resource_alignment= then to make it behave in a more
compatible way then?ÂÂresource_alignment=64k,resource_resize=off?
Thanks,

Alex