Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1411855 > unrolled thread

[RESEND PATCH v2 0/4] PCI: Add support for enforcing all MMIO BARs not to share PAGE_SIZE

Started byYongji Xie <xyjxie@linux.vnet.ibm.com>
First post2016-06-02 08:00 +0200
Last post2016-06-02 08:10 +0200
Articles 3 — 1 participant

Back to article view | Back to linux.kernel


Contents

  [RESEND PATCH v2 0/4] PCI: Add support for enforcing all MMIO BARs not to share PAGE_SIZE Yongji Xie <xyjxie@linux.vnet.ibm.com> - 2016-06-02 08:00 +0200
    [RESEND PATCH v2 4/4] PCI: Add support for enforcing all MMIO BARs to be page aligned Yongji Xie <xyjxie@linux.vnet.ibm.com> - 2016-06-02 08:10 +0200
    [RESEND PATCH v2 1/4] PCI: Ignore resource_alignment if PCI_PROBE_ONLY was set Yongji Xie <xyjxie@linux.vnet.ibm.com> - 2016-06-02 08:10 +0200

#1411855 — [RESEND PATCH v2 0/4] PCI: Add support for enforcing all MMIO BARs not to share PAGE_SIZE

FromYongji Xie <xyjxie@linux.vnet.ibm.com>
Date2016-06-02 08:00 +0200
Subject[RESEND PATCH v2 0/4] PCI: Add support for enforcing all MMIO BARs not to share PAGE_SIZE
Message-ID<rFtPI-1W8-7@gated-at.bofh.it>
This series aims to add an option for PCI resource allocator to  
force BARs not to share PAGE_SIZE. This would make sense to VFIO 
driver. Because current VFIO implementation disallows to mmap 
sub-page(size < PAGE_SIZE) MMIO BARs which may share the same page 
with other BARs for security reasons. Thus, we have to handle mmio 
access to these BARs in QEMU emulation rather than in guest which 
will cause some performance loss.

To achieve that, we would like to make use of the existing 
resource_alignment kernel parameter and force a minimum alignment 
of PAGE_SIZE. It's flexible. And we can default to enable it on  
some archs which may easily hit the performance issue because of  
their 64K page.

In this series, patch 1 fixed a bug of resource_alignment; patch 2,3 
tried to add a new option for resource_alignment to use 
IORESOURCE_STARTALIGN to specify the alignment of PCI BARs; patch 4 
modified resource_alignment to support syntax which can be used to  
enforce the alignment of all MMIO BARs to be at least PAGE_SIZE.

Changelog v2:
- Ignore enforced alignment to VF BARs on pci_reassigndev_resource_alignment()

Yongji Xie (4):
  PCI: Ignore resource_alignment if PCI_PROBE_ONLY was set
  PCI: Do not Use IORESOURCE_STARTALIGN to identify bridge resources
  PCI: Add a new option for resource_alignment to reassign alignment
  PCI: Add support for enforcing all MMIO BARs to be page aligned

 Documentation/kernel-parameters.txt |    7 ++-
 arch/powerpc/include/asm/pci.h      |    2 +
 drivers/pci/pci.c                   |  109 ++++++++++++++++++++++++++++-------
 drivers/pci/setup-bus.c             |    9 ++-
 4 files changed, 102 insertions(+), 25 deletions(-)

-- 
1.7.9.5

[toc] | [next] | [standalone]


#1411859 — [RESEND PATCH v2 4/4] PCI: Add support for enforcing all MMIO BARs to be page aligned

FromYongji Xie <xyjxie@linux.vnet.ibm.com>
Date2016-06-02 08:10 +0200
Subject[RESEND PATCH v2 4/4] PCI: Add support for enforcing all MMIO BARs to be page aligned
Message-ID<rFtZn-2eE-9@gated-at.bofh.it>
In reply to#1411855
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.

This is because vfio will not allow to passthrough one BAR's
mmio page which may be shared with other BARs. Otherwise,
there will be a backdoor that guest can use to access BARs
of other guest.

To solve this issue, this patch modifies resource_alignment
to support syntax where multiple devices get the same
alignment. So we can use something like
"pci=resource_alignment=*:*:*.*:noresize" 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.

And we also define a macro PCIBIOS_MIN_ALIGNMENT to enable this
automatically on PPC64 platform which can easily hit this issue
because its PAGE_SIZE is 64KB.

Note that this would not be applied to VFs whose BARs are always
page aligned and should be never reassigned according to SRIOV
spec.

Signed-off-by: Yongji Xie <xyjxie@linux.vnet.ibm.com>
---
 Documentation/kernel-parameters.txt |    2 ++
 arch/powerpc/include/asm/pci.h      |    2 ++
 drivers/pci/pci.c                   |   68 +++++++++++++++++++++++++++++------
 3 files changed, 61 insertions(+), 11 deletions(-)

diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index c4802f5..cb09503 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -3003,6 +3003,8 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
 				aligned memory resources.
 				If <order of align> is not specified,
 				PAGE_SIZE is used as alignment.
+				<domain>, <bus>, <slot> and <func> can be set to
+				"*" which means match all values.
 				PCI-PCI bridge can be specified, if resource
 				windows need to be expanded.
 				noresize: Don't change the resources' sizes when
diff --git a/arch/powerpc/include/asm/pci.h b/arch/powerpc/include/asm/pci.h
index a6f3ac0..742fd34 100644
--- a/arch/powerpc/include/asm/pci.h
+++ b/arch/powerpc/include/asm/pci.h
@@ -28,6 +28,8 @@
 #define PCIBIOS_MIN_IO		0x1000
 #define PCIBIOS_MIN_MEM		0x10000000
 
+#define PCIBIOS_MIN_ALIGNMENT  PAGE_SIZE
+
 struct pci_dev;
 
 /* Values for the `which' argument to sys_pciconfig_iobase syscall.  */
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 3ee13e5..664f295 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -4759,7 +4759,12 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
 	int seg, bus, slot, func, align_order, count;
 	resource_size_t align = 0;
 	char *p;
+	bool invalid = false;
 
+#ifdef PCIBIOS_MIN_ALIGNMENT
+	align = PCIBIOS_MIN_ALIGNMENT;
+	*resize = false;
+#endif
 	spin_lock(&resource_alignment_lock);
 	p = resource_alignment_param;
 	while (*p) {
@@ -4776,16 +4781,49 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
 		} else {
 			align_order = -1;
 		}
-		if (sscanf(p, "%x:%x:%x.%x%n",
-			&seg, &bus, &slot, &func, &count) != 4) {
+		if (p[0] == '*' && p[1] == ':') {
+			seg = -1;
+			count = 1;
+		} else if (sscanf(p, "%x%n", &seg, &count) != 1 ||
+				p[count] != ':') {
+			invalid = true;
+			break;
+		}
+		p += count + 1;
+		if (*p == '*') {
+			bus = -1;
+			count = 1;
+		} else if (sscanf(p, "%x%n", &bus, &count) != 1) {
+			invalid = true;
+			break;
+		}
+		p += count;
+		if (*p == '.') {
+			slot = bus;
+			bus = seg;
 			seg = 0;
-			if (sscanf(p, "%x:%x.%x%n",
-					&bus, &slot, &func, &count) != 3) {
-				/* Invalid format */
-				printk(KERN_ERR "PCI: Can't parse resource_alignment parameter: %s\n",
-					p);
+			p++;
+		} else if (*p == ':') {
+			p++;
+			if (p[0] == '*' && p[1] == '.') {
+				slot = -1;
+				count = 1;
+			} else if (sscanf(p, "%x%n", &slot, &count) != 1 ||
+					p[count] != '.') {
+				invalid = true;
 				break;
 			}
+			p += count + 1;
+		} else {
+			invalid = true;
+			break;
+		}
+		if (*p == '*') {
+			func = -1;
+			count = 1;
+		} else if (sscanf(p, "%x%n", &func, &count) != 1) {
+			invalid = true;
+			break;
 		}
 		p += count;
 		if (!strncmp(p, ":noresize", 9)) {
@@ -4793,10 +4831,10 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
 			p += 9;
 		} else
 			*resize = true;
-		if (seg == pci_domain_nr(dev->bus) &&
-			bus == dev->bus->number &&
-			slot == PCI_SLOT(dev->devfn) &&
-			func == PCI_FUNC(dev->devfn)) {
+		if ((seg == pci_domain_nr(dev->bus) || seg == -1) &&
+			(bus == dev->bus->number || bus == -1) &&
+			(slot == PCI_SLOT(dev->devfn) || slot == -1) &&
+			(func == PCI_FUNC(dev->devfn) || func == -1)) {
 			if (align_order == -1)
 				align = PAGE_SIZE;
 			else
@@ -4806,10 +4844,14 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev,
 		}
 		if (*p != ';' && *p != ',') {
 			/* End of param or invalid format */
+			invalid = true;
 			break;
 		}
 		p++;
 	}
+	if (invalid)
+		printk(KERN_ERR "PCI: Can't parse resource_alignment parameter:%s\n",
+				p);
 	spin_unlock(&resource_alignment_lock);
 	return align;
 }
@@ -4829,6 +4871,10 @@ void pci_reassigndev_resource_alignment(struct pci_dev *dev)
 	resource_size_t align, size;
 	u16 command;
 
+	/* We should never try to reassign VF's alignment */
+	if (dev->is_virtfn)
+		return;
+
 	/* check if specified PCI is target device to reassign */
 	align = pci_specified_resource_alignment(dev, &resize);
 	if (!align)
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]


#1411861 — [RESEND PATCH v2 1/4] PCI: Ignore resource_alignment if PCI_PROBE_ONLY was set

FromYongji Xie <xyjxie@linux.vnet.ibm.com>
Date2016-06-02 08:10 +0200
Subject[RESEND PATCH v2 1/4] PCI: Ignore resource_alignment if PCI_PROBE_ONLY was set
Message-ID<rFtZn-2eE-17@gated-at.bofh.it>
In reply to#1411855
The resource_alignment will releases memory resources allocated
by firmware so that kernel can reassign new resources later on.
But this will cause the problem that no resources can be
allocated by kernel if PCI_PROBE_ONLY was set, e.g. on pSeries
platform because PCI_PROBE_ONLY force kernel to use firmware
setup and not to reassign any resources.

To solve this problem, this patch ignores resource_alignment
if PCI_PROBE_ONLY was set.

Signed-off-by: Yongji Xie <xyjxie@linux.vnet.ibm.com>
---
 drivers/pci/pci.c |    6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index c8b4dbd..a259394 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -4761,6 +4761,12 @@ static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev)
 	spin_lock(&resource_alignment_lock);
 	p = resource_alignment_param;
 	while (*p) {
+		if (pci_has_flag(PCI_PROBE_ONLY)) {
+			printk(KERN_ERR "PCI: Ignore resource_alignment parameter: %s with PCI_PROBE_ONLY set\n",
+					p);
+			*p = 0;
+			break;
+		}
 		count = 0;
 		if (sscanf(p, "%d%n", &align_order, &count) == 1 &&
 							p[count] == '@') {
-- 
1.7.9.5

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web