2010-07-19 03:16:31

by Alex Williamson

[permalink] [raw]
Subject: [PATCH] pci: Allow read/write access to I/O port resources

PCI sysfs resource files currently only allow mmap'ing, which doesn't
work on all architectures, particularly x86. Add read/write to
allow access to these device regions using sysfs.

Signed-off-by: Alex Williamson <[email protected]>
---

I intend to use this for KVM PCI device assignment to avoid raw in/out
in the code and allow permissions based on pci sysfs files rather than
iopl/ioperm.

Documentation/filesystems/sysfs-pci.txt | 7 ++-
drivers/pci/pci-sysfs.c | 68 +++++++++++++++++++++++++++++++
2 files changed, 73 insertions(+), 2 deletions(-)

diff --git a/Documentation/filesystems/sysfs-pci.txt b/Documentation/filesystems/sysfs-pci.txt
index 85354b3..74eaac2 100644
--- a/Documentation/filesystems/sysfs-pci.txt
+++ b/Documentation/filesystems/sysfs-pci.txt
@@ -39,7 +39,7 @@ files, each with their own function.
local_cpus nearby CPU mask (cpumask, ro)
remove remove device from kernel's list (ascii, wo)
resource PCI resource host addresses (ascii, ro)
- resource0..N PCI resource N, if present (binary, mmap)
+ resource0..N PCI resource N, if present (binary, mmap, rw[1])
resource0_wc..N_wc PCI WC map resource N, if prefetchable (binary, mmap)
rom PCI ROM resource, if present (binary, ro)
subsystem_device PCI subsystem device (ascii, ro)
@@ -54,13 +54,16 @@ files, each with their own function.
binary - file contains binary data
cpumask - file contains a cpumask type

+[1] rw for RESOURCE_IO (I/O port) regions only
+
The read only files are informational, writes to them will be ignored, with
the exception of the 'rom' file. Writable files can be used to perform
actions on the device (e.g. changing config space, detaching a device).
mmapable files are available via an mmap of the file at offset 0 and can be
used to do actual device programming from userspace. Note that some platforms
don't support mmapping of certain resources, so be sure to check the return
-value from any attempted mmap.
+value from any attempted mmap. The most notable of these are I/O port
+resources, which also provide read/write access.

The 'enable' file provides a counter that indicates how many times the device
has been enabled. If the 'enable' file currently returns '4', and a '1' is
diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c
index 5935b85..c98ea15 100644
--- a/drivers/pci/pci-sysfs.c
+++ b/drivers/pci/pci-sysfs.c
@@ -778,6 +778,70 @@ pci_mmap_resource_wc(struct file *filp, struct kobject *kobj,
return pci_mmap_resource(kobj, attr, vma, 1);
}

+static ssize_t
+pci_resource_io(struct file *filp, struct kobject *kobj,
+ struct bin_attribute *attr, char *buf,
+ loff_t off, size_t count, int write)
+{
+ struct pci_dev *pdev = to_pci_dev(container_of(kobj,
+ struct device, kobj));
+ struct resource *res = (struct resource *)attr->private;
+ unsigned long port = off;
+ int i;
+
+ for (i = 0; i < PCI_ROM_RESOURCE; i++)
+ if (res == &pdev->resource[i])
+ break;
+ if (i >= PCI_ROM_RESOURCE)
+ return -ENODEV;
+
+ port += pci_resource_start(pdev, i);
+
+ if (port > pci_resource_end(pdev, i))
+ return 0;
+
+ if (port + count - 1 > pci_resource_end(pdev, i))
+ return -EINVAL;
+
+ switch (count) {
+ case 1:
+ if (write)
+ outb(*(u8 *)buf, port);
+ else
+ *(u8 *)buf = inb(port);
+ return 1;
+ case 2:
+ if (write)
+ outw(*(u16 *)buf, port);
+ else
+ *(u16 *)buf = inw(port);
+ return 2;
+ case 4:
+ if (write)
+ outl(*(u32 *)buf, port);
+ else
+ *(u32 *)buf = inl(port);
+ return 4;
+ }
+ return -EINVAL;
+}
+
+static ssize_t
+pci_read_resource_io(struct file *filp, struct kobject *kobj,
+ struct bin_attribute *attr, char *buf,
+ loff_t off, size_t count)
+{
+ return pci_resource_io(filp, kobj, attr, buf, off, count, 0);
+}
+
+static ssize_t
+pci_write_resource_io(struct file *filp, struct kobject *kobj,
+ struct bin_attribute *attr, char *buf,
+ loff_t off, size_t count)
+{
+ return pci_resource_io(filp, kobj, attr, buf, off, count, 1);
+}
+
/**
* pci_remove_resource_files - cleanup resource files
* @pdev: dev to cleanup
@@ -828,6 +892,10 @@ static int pci_create_attr(struct pci_dev *pdev, int num, int write_combine)
sprintf(res_attr_name, "resource%d", num);
res_attr->mmap = pci_mmap_resource_uc;
}
+ if (pci_resource_flags(pdev, num) & IORESOURCE_IO) {
+ res_attr->read = pci_read_resource_io;
+ res_attr->write = pci_write_resource_io;
+ }
res_attr->attr.name = res_attr_name;
res_attr->attr.mode = S_IRUSR | S_IWUSR;
res_attr->size = pci_resource_len(pdev, num);


2010-07-19 05:57:48

by Rolf Eike Beer

[permalink] [raw]
Subject: Re: [PATCH] pci: Allow read/write access to I/O port resources

Alex Williamson wrote:
> PCI sysfs resource files currently only allow mmap'ing, which doesn't
> work on all architectures, particularly x86. Add read/write to
> allow access to these device regions using sysfs.

I think you should update that description a bit. It took me some time to find
out you are not talking about mmap() on PCI resource files in general, but
those that are backed by I/O BARs. The other ones work fine on x86 with mmap().

> +static ssize_t
> +pci_resource_io(struct file *filp, struct kobject *kobj,
> + struct bin_attribute *attr, char *buf,
> + loff_t off, size_t count, int write)
> +{

The "write" can be bool.

> + struct pci_dev *pdev = to_pci_dev(container_of(kobj,
> + struct device,
kobj));
> + struct resource *res = (struct resource *)attr->private;

No need to cast here, private is void* anyway.

Eike


Attachments:
signature.asc (198.00 B)
This is a digitally signed message part.

2010-07-19 15:46:21

by Alex Williamson

[permalink] [raw]
Subject: [PATCH v2] pci: Allow read/write access to sysfs I/O port resources

PCI sysfs resource files currently only allow mmap'ing. On x86 this
works fine for memory backed BARs, but doesn't work at all for I/O
port backed BARs. Add read/write to I/O port PCI sysfs resource
files to allow userspace access to these device regions.

Signed-off-by: Alex Williamson <[email protected]>
---

v2: hopefully better commit log, remove unneeded cast, use bool

Documentation/filesystems/sysfs-pci.txt | 7 ++-
drivers/pci/pci-sysfs.c | 68 +++++++++++++++++++++++++++++++
2 files changed, 73 insertions(+), 2 deletions(-)

diff --git a/Documentation/filesystems/sysfs-pci.txt b/Documentation/filesystems/sysfs-pci.txt
index 85354b3..74eaac2 100644
--- a/Documentation/filesystems/sysfs-pci.txt
+++ b/Documentation/filesystems/sysfs-pci.txt
@@ -39,7 +39,7 @@ files, each with their own function.
local_cpus nearby CPU mask (cpumask, ro)
remove remove device from kernel's list (ascii, wo)
resource PCI resource host addresses (ascii, ro)
- resource0..N PCI resource N, if present (binary, mmap)
+ resource0..N PCI resource N, if present (binary, mmap, rw[1])
resource0_wc..N_wc PCI WC map resource N, if prefetchable (binary, mmap)
rom PCI ROM resource, if present (binary, ro)
subsystem_device PCI subsystem device (ascii, ro)
@@ -54,13 +54,16 @@ files, each with their own function.
binary - file contains binary data
cpumask - file contains a cpumask type

+[1] rw for RESOURCE_IO (I/O port) regions only
+
The read only files are informational, writes to them will be ignored, with
the exception of the 'rom' file. Writable files can be used to perform
actions on the device (e.g. changing config space, detaching a device).
mmapable files are available via an mmap of the file at offset 0 and can be
used to do actual device programming from userspace. Note that some platforms
don't support mmapping of certain resources, so be sure to check the return
-value from any attempted mmap.
+value from any attempted mmap. The most notable of these are I/O port
+resources, which also provide read/write access.

The 'enable' file provides a counter that indicates how many times the device
has been enabled. If the 'enable' file currently returns '4', and a '1' is
diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c
index 5935b85..f7692dc 100644
--- a/drivers/pci/pci-sysfs.c
+++ b/drivers/pci/pci-sysfs.c
@@ -778,6 +778,70 @@ pci_mmap_resource_wc(struct file *filp, struct kobject *kobj,
return pci_mmap_resource(kobj, attr, vma, 1);
}

+static ssize_t
+pci_resource_io(struct file *filp, struct kobject *kobj,
+ struct bin_attribute *attr, char *buf,
+ loff_t off, size_t count, bool write)
+{
+ struct pci_dev *pdev = to_pci_dev(container_of(kobj,
+ struct device, kobj));
+ struct resource *res = attr->private;
+ unsigned long port = off;
+ int i;
+
+ for (i = 0; i < PCI_ROM_RESOURCE; i++)
+ if (res == &pdev->resource[i])
+ break;
+ if (i >= PCI_ROM_RESOURCE)
+ return -ENODEV;
+
+ port += pci_resource_start(pdev, i);
+
+ if (port > pci_resource_end(pdev, i))
+ return 0;
+
+ if (port + count - 1 > pci_resource_end(pdev, i))
+ return -EINVAL;
+
+ switch (count) {
+ case 1:
+ if (write)
+ outb(*(u8 *)buf, port);
+ else
+ *(u8 *)buf = inb(port);
+ return 1;
+ case 2:
+ if (write)
+ outw(*(u16 *)buf, port);
+ else
+ *(u16 *)buf = inw(port);
+ return 2;
+ case 4:
+ if (write)
+ outl(*(u32 *)buf, port);
+ else
+ *(u32 *)buf = inl(port);
+ return 4;
+ }
+ return -EINVAL;
+}
+
+static ssize_t
+pci_read_resource_io(struct file *filp, struct kobject *kobj,
+ struct bin_attribute *attr, char *buf,
+ loff_t off, size_t count)
+{
+ return pci_resource_io(filp, kobj, attr, buf, off, count, false);
+}
+
+static ssize_t
+pci_write_resource_io(struct file *filp, struct kobject *kobj,
+ struct bin_attribute *attr, char *buf,
+ loff_t off, size_t count)
+{
+ return pci_resource_io(filp, kobj, attr, buf, off, count, true);
+}
+
/**
* pci_remove_resource_files - cleanup resource files
* @pdev: dev to cleanup
@@ -828,6 +892,10 @@ static int pci_create_attr(struct pci_dev *pdev, int num, int write_combine)
sprintf(res_attr_name, "resource%d", num);
res_attr->mmap = pci_mmap_resource_uc;
}
+ if (pci_resource_flags(pdev, num) & IORESOURCE_IO) {
+ res_attr->read = pci_read_resource_io;
+ res_attr->write = pci_write_resource_io;
+ }
res_attr->attr.name = res_attr_name;
res_attr->attr.mode = S_IRUSR | S_IWUSR;
res_attr->size = pci_resource_len(pdev, num);

2010-07-23 15:11:20

by Alex Williamson

[permalink] [raw]
Subject: Re: [PATCH v2] pci: Allow read/write access to sysfs I/O port resources

On Mon, 2010-07-19 at 09:45 -0600, Alex Williamson wrote:
> PCI sysfs resource files currently only allow mmap'ing. On x86 this
> works fine for memory backed BARs, but doesn't work at all for I/O
> port backed BARs. Add read/write to I/O port PCI sysfs resource
> files to allow userspace access to these device regions.
>
> Signed-off-by: Alex Williamson <[email protected]>
> ---

Any thoughts on this for -next? Thanks,

Alex

> v2: hopefully better commit log, remove unneeded cast, use bool
>
> Documentation/filesystems/sysfs-pci.txt | 7 ++-
> drivers/pci/pci-sysfs.c | 68 +++++++++++++++++++++++++++++++
> 2 files changed, 73 insertions(+), 2 deletions(-)
>
> diff --git a/Documentation/filesystems/sysfs-pci.txt b/Documentation/filesystems/sysfs-pci.txt
> index 85354b3..74eaac2 100644
> --- a/Documentation/filesystems/sysfs-pci.txt
> +++ b/Documentation/filesystems/sysfs-pci.txt
> @@ -39,7 +39,7 @@ files, each with their own function.
> local_cpus nearby CPU mask (cpumask, ro)
> remove remove device from kernel's list (ascii, wo)
> resource PCI resource host addresses (ascii, ro)
> - resource0..N PCI resource N, if present (binary, mmap)
> + resource0..N PCI resource N, if present (binary, mmap, rw[1])
> resource0_wc..N_wc PCI WC map resource N, if prefetchable (binary, mmap)
> rom PCI ROM resource, if present (binary, ro)
> subsystem_device PCI subsystem device (ascii, ro)
> @@ -54,13 +54,16 @@ files, each with their own function.
> binary - file contains binary data
> cpumask - file contains a cpumask type
>
> +[1] rw for RESOURCE_IO (I/O port) regions only
> +
> The read only files are informational, writes to them will be ignored, with
> the exception of the 'rom' file. Writable files can be used to perform
> actions on the device (e.g. changing config space, detaching a device).
> mmapable files are available via an mmap of the file at offset 0 and can be
> used to do actual device programming from userspace. Note that some platforms
> don't support mmapping of certain resources, so be sure to check the return
> -value from any attempted mmap.
> +value from any attempted mmap. The most notable of these are I/O port
> +resources, which also provide read/write access.
>
> The 'enable' file provides a counter that indicates how many times the device
> has been enabled. If the 'enable' file currently returns '4', and a '1' is
> diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c
> index 5935b85..f7692dc 100644
> --- a/drivers/pci/pci-sysfs.c
> +++ b/drivers/pci/pci-sysfs.c
> @@ -778,6 +778,70 @@ pci_mmap_resource_wc(struct file *filp, struct kobject *kobj,
> return pci_mmap_resource(kobj, attr, vma, 1);
> }
>
> +static ssize_t
> +pci_resource_io(struct file *filp, struct kobject *kobj,
> + struct bin_attribute *attr, char *buf,
> + loff_t off, size_t count, bool write)
> +{
> + struct pci_dev *pdev = to_pci_dev(container_of(kobj,
> + struct device, kobj));
> + struct resource *res = attr->private;
> + unsigned long port = off;
> + int i;
> +
> + for (i = 0; i < PCI_ROM_RESOURCE; i++)
> + if (res == &pdev->resource[i])
> + break;
> + if (i >= PCI_ROM_RESOURCE)
> + return -ENODEV;
> +
> + port += pci_resource_start(pdev, i);
> +
> + if (port > pci_resource_end(pdev, i))
> + return 0;
> +
> + if (port + count - 1 > pci_resource_end(pdev, i))
> + return -EINVAL;
> +
> + switch (count) {
> + case 1:
> + if (write)
> + outb(*(u8 *)buf, port);
> + else
> + *(u8 *)buf = inb(port);
> + return 1;
> + case 2:
> + if (write)
> + outw(*(u16 *)buf, port);
> + else
> + *(u16 *)buf = inw(port);
> + return 2;
> + case 4:
> + if (write)
> + outl(*(u32 *)buf, port);
> + else
> + *(u32 *)buf = inl(port);
> + return 4;
> + }
> + return -EINVAL;
> +}
> +
> +static ssize_t
> +pci_read_resource_io(struct file *filp, struct kobject *kobj,
> + struct bin_attribute *attr, char *buf,
> + loff_t off, size_t count)
> +{
> + return pci_resource_io(filp, kobj, attr, buf, off, count, false);
> +}
> +
> +static ssize_t
> +pci_write_resource_io(struct file *filp, struct kobject *kobj,
> + struct bin_attribute *attr, char *buf,
> + loff_t off, size_t count)
> +{
> + return pci_resource_io(filp, kobj, attr, buf, off, count, true);
> +}
> +
> /**
> * pci_remove_resource_files - cleanup resource files
> * @pdev: dev to cleanup
> @@ -828,6 +892,10 @@ static int pci_create_attr(struct pci_dev *pdev, int num, int write_combine)
> sprintf(res_attr_name, "resource%d", num);
> res_attr->mmap = pci_mmap_resource_uc;
> }
> + if (pci_resource_flags(pdev, num) & IORESOURCE_IO) {
> + res_attr->read = pci_read_resource_io;
> + res_attr->write = pci_write_resource_io;
> + }
> res_attr->attr.name = res_attr_name;
> res_attr->attr.mode = S_IRUSR | S_IWUSR;
> res_attr->size = pci_resource_len(pdev, num);
>


2010-07-23 16:27:54

by Jesse Barnes

[permalink] [raw]
Subject: Re: [PATCH v2] pci: Allow read/write access to sysfs I/O port resources

On Fri, 23 Jul 2010 09:10:27 -0600
Alex Williamson <[email protected]> wrote:

> On Mon, 2010-07-19 at 09:45 -0600, Alex Williamson wrote:
> > PCI sysfs resource files currently only allow mmap'ing. On x86 this
> > works fine for memory backed BARs, but doesn't work at all for I/O
> > port backed BARs. Add read/write to I/O port PCI sysfs resource
> > files to allow userspace access to these device regions.
> >
> > Signed-off-by: Alex Williamson <[email protected]>
> > ---
>
> Any thoughts on this for -next? Thanks,

Yeah, I like it. Unless there are objections I'll pull it in.

Thanks,
--
Jesse Barnes, Intel Open Source Technology Center

2010-07-23 17:43:51

by Chris Wright

[permalink] [raw]
Subject: Re: [PATCH v2] pci: Allow read/write access to sysfs I/O port resources

* Jesse Barnes ([email protected]) wrote:
> On Fri, 23 Jul 2010 09:10:27 -0600
> Alex Williamson <[email protected]> wrote:
>
> > On Mon, 2010-07-19 at 09:45 -0600, Alex Williamson wrote:
> > > PCI sysfs resource files currently only allow mmap'ing. On x86 this
> > > works fine for memory backed BARs, but doesn't work at all for I/O
> > > port backed BARs. Add read/write to I/O port PCI sysfs resource
> > > files to allow userspace access to these device regions.
> > >
> > > Signed-off-by: Alex Williamson <[email protected]>
> > > ---
> >
> > Any thoughts on this for -next? Thanks,
>
> Yeah, I like it. Unless there are objections I'll pull it in.

Acked-by: Chris Wright <[email protected]>

I think we may have some room for follow-on cleanups so consolidate the
ioport read/write functions w/ other HAVE_PCI_LEGACY bits...

thanks,
-chris

2010-07-30 16:36:23

by Jesse Barnes

[permalink] [raw]
Subject: Re: [PATCH v2] pci: Allow read/write access to sysfs I/O port resources

On Mon, 19 Jul 2010 09:45:34 -0600
Alex Williamson <[email protected]> wrote:

> PCI sysfs resource files currently only allow mmap'ing. On x86 this
> works fine for memory backed BARs, but doesn't work at all for I/O
> port backed BARs. Add read/write to I/O port PCI sysfs resource
> files to allow userspace access to these device regions.
>
> Signed-off-by: Alex Williamson <[email protected]>

Applied to linux-next, thanks.

--
Jesse Barnes, Intel Open Source Technology Center