Sometimes, we want to control which of the matching drivers
binds to a subchannel device (e.g. for subchannels we want to
handle via vfio-ccw).
For pci devices, a mechanism to do so has been introduced in
782a985d7af2 ("PCI: Introduce new device binding path using
pci_dev.driver_override"). It makes sense to introduce the
driver_override attribute for subchannel devices as well, so
that we can easily extend the 'driverctl' tool (which makes
use of the driver_override attribute for pci).
Note that unlike pci we still require a driver override to
match the subchannel type; matching more than one subchannel
type is probably not useful anyway.
Signed-off-by: Cornelia Huck <[email protected]>
---
Lightly tested; did not yet attempt to adapt driverctl to actually
make use of it.
For some background, refer to the thread around
https://lore.kernel.org/kvm/[email protected]/
---
Documentation/ABI/testing/sysfs-bus-css | 23 +++++++++++
drivers/s390/cio/cio.h | 1 +
drivers/s390/cio/css.c | 53 +++++++++++++++++++++++++
3 files changed, 77 insertions(+)
diff --git a/Documentation/ABI/testing/sysfs-bus-css b/Documentation/ABI/testing/sysfs-bus-css
index 2979c40c10e9..966f8504bd7b 100644
--- a/Documentation/ABI/testing/sysfs-bus-css
+++ b/Documentation/ABI/testing/sysfs-bus-css
@@ -33,3 +33,26 @@ Description: Contains the PIM/PAM/POM values, as reported by the
in sync with the values current in the channel subsystem).
Note: This is an I/O-subchannel specific attribute.
Users: s390-tools, HAL
+
+What: /sys/bus/css/devices/.../driver_override
+Date: June 2019
+Contact: Cornelia Huck <[email protected]>
+ [email protected]
+Description: This file allows the driver for a device to be specified. When
+ specified, only a driver with a name matching the value written
+ to driver_override will have an opportunity to bind to the
+ device. The override is specified by writing a string to the
+ driver_override file (echo vfio-ccw > driver_override) and
+ may be cleared with an empty string (echo > driver_override).
+ This returns the device to standard matching rules binding.
+ Writing to driver_override does not automatically unbind the
+ device from its current driver or make any attempt to
+ automatically load the specified driver. If no driver with a
+ matching name is currently loaded in the kernel, the device
+ will not bind to any driver. This also allows devices to
+ opt-out of driver binding using a driver_override name such as
+ "none". Only a single driver may be specified in the override,
+ there is no support for parsing delimiters.
+ Note that unlike the mechanism of the same name for pci, this
+ file does not allow to override basic matching rules. I.e.,
+ the driver must still match the subchannel type of the device.
diff --git a/drivers/s390/cio/cio.h b/drivers/s390/cio/cio.h
index 06a91743335a..8c4af88f1ac3 100644
--- a/drivers/s390/cio/cio.h
+++ b/drivers/s390/cio/cio.h
@@ -113,6 +113,7 @@ struct subchannel {
enum sch_todo todo;
struct work_struct todo_work;
struct schib_config config;
+ char *driver_override; /* Driver name to force a match */
} __attribute__ ((aligned(8)));
DECLARE_PER_CPU_ALIGNED(struct irb, cio_irb);
diff --git a/drivers/s390/cio/css.c b/drivers/s390/cio/css.c
index aea502922646..f3436a17e3b5 100644
--- a/drivers/s390/cio/css.c
+++ b/drivers/s390/cio/css.c
@@ -165,6 +165,7 @@ static void css_subchannel_release(struct device *dev)
sch->config.intparm = 0;
cio_commit_config(sch);
+ kfree(sch->driver_override);
kfree(sch->lock);
kfree(sch);
}
@@ -315,9 +316,57 @@ static ssize_t modalias_show(struct device *dev, struct device_attribute *attr,
static DEVICE_ATTR_RO(modalias);
+static ssize_t driver_override_store(struct device *dev,
+ struct device_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct subchannel *sch = to_subchannel(dev);
+ char *driver_override, *old, *cp;
+
+ /* We need to keep extra room for a newline */
+ if (count >= (PAGE_SIZE - 1))
+ return -EINVAL;
+
+ driver_override = kstrndup(buf, count, GFP_KERNEL);
+ if (!driver_override)
+ return -ENOMEM;
+
+ cp = strchr(driver_override, '\n');
+ if (cp)
+ *cp = '\0';
+
+ device_lock(dev);
+ old = sch->driver_override;
+ if (strlen(driver_override)) {
+ sch->driver_override = driver_override;
+ } else {
+ kfree(driver_override);
+ sch->driver_override = NULL;
+ }
+ device_unlock(dev);
+
+ kfree(old);
+
+ return count;
+}
+
+static ssize_t driver_override_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ struct subchannel *sch = to_subchannel(dev);
+ ssize_t len;
+
+ device_lock(dev);
+ len = snprintf(buf, PAGE_SIZE, "%s\n", sch->driver_override);
+ device_unlock(dev);
+ return len;
+}
+static DEVICE_ATTR_RW(driver_override);
+
static struct attribute *subch_attrs[] = {
&dev_attr_type.attr,
&dev_attr_modalias.attr,
+ &dev_attr_driver_override.attr,
NULL,
};
@@ -1222,6 +1271,10 @@ static int css_bus_match(struct device *dev, struct device_driver *drv)
struct css_driver *driver = to_cssdriver(drv);
struct css_device_id *id;
+ /* When driver_override is set, only bind to the matching driver */
+ if (sch->driver_override && strcmp(sch->driver_override, drv->name))
+ return 0;
+
for (id = driver->subchannel_type; id->match_flags; id++) {
if (sch->st == id->type)
return 1;
--
2.20.1
On Thu, 13 Jun 2019 13:08:15 +0200
Cornelia Huck <[email protected]> wrote:
> Sometimes, we want to control which of the matching drivers
> binds to a subchannel device (e.g. for subchannels we want to
> handle via vfio-ccw).
>
> For pci devices, a mechanism to do so has been introduced in
> 782a985d7af2 ("PCI: Introduce new device binding path using
> pci_dev.driver_override"). It makes sense to introduce the
> driver_override attribute for subchannel devices as well, so
> that we can easily extend the 'driverctl' tool (which makes
> use of the driver_override attribute for pci).
>
> Note that unlike pci we still require a driver override to
> match the subchannel type; matching more than one subchannel
> type is probably not useful anyway.
>
> Signed-off-by: Cornelia Huck <[email protected]>
> ---
>
> Lightly tested; did not yet attempt to adapt driverctl to actually
> make use of it.
Friendly ping.
In the meanwhile, I figured out that you do not need to adapt driverctl
at all, but just need to pass it '-b css' to work on the css bus; this
seems to work just fine with this patch applied.
>
> For some background, refer to the thread around
> https://lore.kernel.org/kvm/[email protected]/
>
> ---
> Documentation/ABI/testing/sysfs-bus-css | 23 +++++++++++
> drivers/s390/cio/cio.h | 1 +
> drivers/s390/cio/css.c | 53 +++++++++++++++++++++++++
> 3 files changed, 77 insertions(+)
>
> diff --git a/Documentation/ABI/testing/sysfs-bus-css b/Documentation/ABI/testing/sysfs-bus-css
> index 2979c40c10e9..966f8504bd7b 100644
> --- a/Documentation/ABI/testing/sysfs-bus-css
> +++ b/Documentation/ABI/testing/sysfs-bus-css
> @@ -33,3 +33,26 @@ Description: Contains the PIM/PAM/POM values, as reported by the
> in sync with the values current in the channel subsystem).
> Note: This is an I/O-subchannel specific attribute.
> Users: s390-tools, HAL
> +
> +What: /sys/bus/css/devices/.../driver_override
> +Date: June 2019
> +Contact: Cornelia Huck <[email protected]>
> + [email protected]
> +Description: This file allows the driver for a device to be specified. When
> + specified, only a driver with a name matching the value written
> + to driver_override will have an opportunity to bind to the
> + device. The override is specified by writing a string to the
> + driver_override file (echo vfio-ccw > driver_override) and
> + may be cleared with an empty string (echo > driver_override).
> + This returns the device to standard matching rules binding.
> + Writing to driver_override does not automatically unbind the
> + device from its current driver or make any attempt to
> + automatically load the specified driver. If no driver with a
> + matching name is currently loaded in the kernel, the device
> + will not bind to any driver. This also allows devices to
> + opt-out of driver binding using a driver_override name such as
> + "none". Only a single driver may be specified in the override,
> + there is no support for parsing delimiters.
> + Note that unlike the mechanism of the same name for pci, this
> + file does not allow to override basic matching rules. I.e.,
> + the driver must still match the subchannel type of the device.
> diff --git a/drivers/s390/cio/cio.h b/drivers/s390/cio/cio.h
> index 06a91743335a..8c4af88f1ac3 100644
> --- a/drivers/s390/cio/cio.h
> +++ b/drivers/s390/cio/cio.h
> @@ -113,6 +113,7 @@ struct subchannel {
> enum sch_todo todo;
> struct work_struct todo_work;
> struct schib_config config;
> + char *driver_override; /* Driver name to force a match */
> } __attribute__ ((aligned(8)));
>
> DECLARE_PER_CPU_ALIGNED(struct irb, cio_irb);
> diff --git a/drivers/s390/cio/css.c b/drivers/s390/cio/css.c
> index aea502922646..f3436a17e3b5 100644
> --- a/drivers/s390/cio/css.c
> +++ b/drivers/s390/cio/css.c
> @@ -165,6 +165,7 @@ static void css_subchannel_release(struct device *dev)
>
> sch->config.intparm = 0;
> cio_commit_config(sch);
> + kfree(sch->driver_override);
> kfree(sch->lock);
> kfree(sch);
> }
> @@ -315,9 +316,57 @@ static ssize_t modalias_show(struct device *dev, struct device_attribute *attr,
>
> static DEVICE_ATTR_RO(modalias);
>
> +static ssize_t driver_override_store(struct device *dev,
> + struct device_attribute *attr,
> + const char *buf, size_t count)
> +{
> + struct subchannel *sch = to_subchannel(dev);
> + char *driver_override, *old, *cp;
> +
> + /* We need to keep extra room for a newline */
> + if (count >= (PAGE_SIZE - 1))
> + return -EINVAL;
> +
> + driver_override = kstrndup(buf, count, GFP_KERNEL);
> + if (!driver_override)
> + return -ENOMEM;
> +
> + cp = strchr(driver_override, '\n');
> + if (cp)
> + *cp = '\0';
> +
> + device_lock(dev);
> + old = sch->driver_override;
> + if (strlen(driver_override)) {
> + sch->driver_override = driver_override;
> + } else {
> + kfree(driver_override);
> + sch->driver_override = NULL;
> + }
> + device_unlock(dev);
> +
> + kfree(old);
> +
> + return count;
> +}
> +
> +static ssize_t driver_override_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct subchannel *sch = to_subchannel(dev);
> + ssize_t len;
> +
> + device_lock(dev);
> + len = snprintf(buf, PAGE_SIZE, "%s\n", sch->driver_override);
> + device_unlock(dev);
> + return len;
> +}
> +static DEVICE_ATTR_RW(driver_override);
> +
> static struct attribute *subch_attrs[] = {
> &dev_attr_type.attr,
> &dev_attr_modalias.attr,
> + &dev_attr_driver_override.attr,
> NULL,
> };
>
> @@ -1222,6 +1271,10 @@ static int css_bus_match(struct device *dev, struct device_driver *drv)
> struct css_driver *driver = to_cssdriver(drv);
> struct css_device_id *id;
>
> + /* When driver_override is set, only bind to the matching driver */
> + if (sch->driver_override && strcmp(sch->driver_override, drv->name))
> + return 0;
> +
> for (id = driver->subchannel_type; id->match_flags; id++) {
> if (sch->st == id->type)
> return 1;
On Fri, 21 Jun 2019 11:56:04 +0200
Cornelia Huck <[email protected]> wrote:
> On Thu, 13 Jun 2019 13:08:15 +0200
> Cornelia Huck <[email protected]> wrote:
>
> > Sometimes, we want to control which of the matching drivers
> > binds to a subchannel device (e.g. for subchannels we want to
> > handle via vfio-ccw).
> >
> > For pci devices, a mechanism to do so has been introduced in
> > 782a985d7af2 ("PCI: Introduce new device binding path using
> > pci_dev.driver_override"). It makes sense to introduce the
> > driver_override attribute for subchannel devices as well, so
> > that we can easily extend the 'driverctl' tool (which makes
> > use of the driver_override attribute for pci).
> >
> > Note that unlike pci we still require a driver override to
> > match the subchannel type; matching more than one subchannel
> > type is probably not useful anyway.
> >
> > Signed-off-by: Cornelia Huck <[email protected]>
Reviewed-by: Halil Pasic <[email protected]>
I guess the '\n' handling is customary, and is what the same what
the pci counterpart (782a985d7af2) does anyway. It bothers
me a little that you don't necessarily get back from with show
what you stored. E.g. # echo -e "bug\nfree"
> /sys/bus/css/devices/0.0.0001/driver_override # echo $?
0
# cat /sys/bus/css/devices/0.0.0001/driver_override
bug
# echo $?
0
But given the previous art (782a985d7af2) I think it is the best way
to do it.
The rest is very straightforward.
> > ---
> >
> > Lightly tested; did not yet attempt to adapt driverctl to actually
> > make use of it.
>
> Friendly ping.
>
> In the meanwhile, I figured out that you do not need to adapt driverctl
> at all, but just need to pass it '-b css' to work on the css bus; this
> seems to work just fine with this patch applied.
>
Interesting. I hope to get around and have a closer look at it
eventually.
Regards,
Halil
On Fri, 21 Jun 2019 15:37:11 +0200
Halil Pasic <[email protected]> wrote:
> On Fri, 21 Jun 2019 11:56:04 +0200
> Cornelia Huck <[email protected]> wrote:
>
> > On Thu, 13 Jun 2019 13:08:15 +0200
> > Cornelia Huck <[email protected]> wrote:
> >
> > > Sometimes, we want to control which of the matching drivers
> > > binds to a subchannel device (e.g. for subchannels we want to
> > > handle via vfio-ccw).
> > >
> > > For pci devices, a mechanism to do so has been introduced in
> > > 782a985d7af2 ("PCI: Introduce new device binding path using
> > > pci_dev.driver_override"). It makes sense to introduce the
> > > driver_override attribute for subchannel devices as well, so
> > > that we can easily extend the 'driverctl' tool (which makes
> > > use of the driver_override attribute for pci).
> > >
> > > Note that unlike pci we still require a driver override to
> > > match the subchannel type; matching more than one subchannel
> > > type is probably not useful anyway.
> > >
> > > Signed-off-by: Cornelia Huck <[email protected]>
>
> Reviewed-by: Halil Pasic <[email protected]>
Thanks!
>
> I guess the '\n' handling is customary, and is what the same what
> the pci counterpart (782a985d7af2) does anyway. It bothers
> me a little that you don't necessarily get back from with show
> what you stored. E.g. # echo -e "bug\nfree"
> > /sys/bus/css/devices/0.0.0001/driver_override # echo $?
> 0
> # cat /sys/bus/css/devices/0.0.0001/driver_override
> bug
> # echo $?
> 0
> But given the previous art (782a985d7af2) I think it is the best way
> to do it.
Yeah; I shamelessly copied a lot :) And I agree it's probably useful to
have both act in the same way.
>
> The rest is very straightforward.
>
> > > ---
> > >
> > > Lightly tested; did not yet attempt to adapt driverctl to actually
> > > make use of it.
> >
> > Friendly ping.
> >
> > In the meanwhile, I figured out that you do not need to adapt driverctl
> > at all, but just need to pass it '-b css' to work on the css bus; this
> > seems to work just fine with this patch applied.
> >
>
> Interesting. I hope to get around and have a closer look at it
> eventually.
Nice; somebody else looking at it is always useful :)
On Thu, 13 Jun 2019, Cornelia Huck wrote:
> Sometimes, we want to control which of the matching drivers
> binds to a subchannel device (e.g. for subchannels we want to
> handle via vfio-ccw).
>
> For pci devices, a mechanism to do so has been introduced in
> 782a985d7af2 ("PCI: Introduce new device binding path using
> pci_dev.driver_override"). It makes sense to introduce the
> driver_override attribute for subchannel devices as well, so
> that we can easily extend the 'driverctl' tool (which makes
> use of the driver_override attribute for pci).
>
> Note that unlike pci we still require a driver override to
> match the subchannel type; matching more than one subchannel
> type is probably not useful anyway.
>
> Signed-off-by: Cornelia Huck <[email protected]>
Reviewed-by: Sebastian Ott <[email protected]>
Should I take that via our git tree or do you have other patches depending
on this one?
On Fri, 21 Jun 2019 18:19:36 +0200 (CEST)
Sebastian Ott <[email protected]> wrote:
> On Thu, 13 Jun 2019, Cornelia Huck wrote:
> > Sometimes, we want to control which of the matching drivers
> > binds to a subchannel device (e.g. for subchannels we want to
> > handle via vfio-ccw).
> >
> > For pci devices, a mechanism to do so has been introduced in
> > 782a985d7af2 ("PCI: Introduce new device binding path using
> > pci_dev.driver_override"). It makes sense to introduce the
> > driver_override attribute for subchannel devices as well, so
> > that we can easily extend the 'driverctl' tool (which makes
> > use of the driver_override attribute for pci).
> >
> > Note that unlike pci we still require a driver override to
> > match the subchannel type; matching more than one subchannel
> > type is probably not useful anyway.
> >
> > Signed-off-by: Cornelia Huck <[email protected]>
>
> Reviewed-by: Sebastian Ott <[email protected]>
>
> Should I take that via our git tree or do you have other patches depending
> on this one?
>
No, this patch is stand-alone; everything else is happening in user
space, so taking it via your tree would be great. Thanks!
On Mon, 24 Jun 2019 09:17:40 +0200
Cornelia Huck <[email protected]> wrote:
> On Fri, 21 Jun 2019 18:19:36 +0200 (CEST)
> Sebastian Ott <[email protected]> wrote:
>
> > On Thu, 13 Jun 2019, Cornelia Huck wrote:
> > > Sometimes, we want to control which of the matching drivers
> > > binds to a subchannel device (e.g. for subchannels we want to
> > > handle via vfio-ccw).
> > >
> > > For pci devices, a mechanism to do so has been introduced in
> > > 782a985d7af2 ("PCI: Introduce new device binding path using
> > > pci_dev.driver_override"). It makes sense to introduce the
> > > driver_override attribute for subchannel devices as well, so
> > > that we can easily extend the 'driverctl' tool (which makes
> > > use of the driver_override attribute for pci).
> > >
> > > Note that unlike pci we still require a driver override to
> > > match the subchannel type; matching more than one subchannel
> > > type is probably not useful anyway.
> > >
> > > Signed-off-by: Cornelia Huck <[email protected]>
> >
> > Reviewed-by: Sebastian Ott <[email protected]>
> >
> > Should I take that via our git tree or do you have other patches depending
> > on this one?
> >
>
> No, this patch is stand-alone; everything else is happening in user
> space, so taking it via your tree would be great. Thanks!
Friendly ping (I don't see it on s390/features yet; or is this going
via some other path?)
On Wed, Jul 03, 2019 at 04:55:06PM +0200, Cornelia Huck wrote:
> On Mon, 24 Jun 2019 09:17:40 +0200
> Cornelia Huck <[email protected]> wrote:
>
> > On Fri, 21 Jun 2019 18:19:36 +0200 (CEST)
> > Sebastian Ott <[email protected]> wrote:
> >
> > > On Thu, 13 Jun 2019, Cornelia Huck wrote:
> > > > Sometimes, we want to control which of the matching drivers
> > > > binds to a subchannel device (e.g. for subchannels we want to
> > > > handle via vfio-ccw).
> > > >
> > > > For pci devices, a mechanism to do so has been introduced in
> > > > 782a985d7af2 ("PCI: Introduce new device binding path using
> > > > pci_dev.driver_override"). It makes sense to introduce the
> > > > driver_override attribute for subchannel devices as well, so
> > > > that we can easily extend the 'driverctl' tool (which makes
> > > > use of the driver_override attribute for pci).
> > > >
> > > > Note that unlike pci we still require a driver override to
> > > > match the subchannel type; matching more than one subchannel
> > > > type is probably not useful anyway.
> > > >
> > > > Signed-off-by: Cornelia Huck <[email protected]>
> > >
> > > Reviewed-by: Sebastian Ott <[email protected]>
> > >
> > > Should I take that via our git tree or do you have other patches depending
> > > on this one?
> > >
> >
> > No, this patch is stand-alone; everything else is happening in user
> > space, so taking it via your tree would be great. Thanks!
>
> Friendly ping (I don't see it on s390/features yet; or is this going
> via some other path?)
>
It is there now. Just waited until latest ci results to push the branch.
Sorry for delay.
On Wed, 3 Jul 2019 17:09:53 +0200
Vasily Gorbik <[email protected]> wrote:
> On Wed, Jul 03, 2019 at 04:55:06PM +0200, Cornelia Huck wrote:
> > On Mon, 24 Jun 2019 09:17:40 +0200
> > Cornelia Huck <[email protected]> wrote:
> >
> > > On Fri, 21 Jun 2019 18:19:36 +0200 (CEST)
> > > Sebastian Ott <[email protected]> wrote:
> > >
> > > > On Thu, 13 Jun 2019, Cornelia Huck wrote:
> > > > > Sometimes, we want to control which of the matching drivers
> > > > > binds to a subchannel device (e.g. for subchannels we want to
> > > > > handle via vfio-ccw).
> > > > >
> > > > > For pci devices, a mechanism to do so has been introduced in
> > > > > 782a985d7af2 ("PCI: Introduce new device binding path using
> > > > > pci_dev.driver_override"). It makes sense to introduce the
> > > > > driver_override attribute for subchannel devices as well, so
> > > > > that we can easily extend the 'driverctl' tool (which makes
> > > > > use of the driver_override attribute for pci).
> > > > >
> > > > > Note that unlike pci we still require a driver override to
> > > > > match the subchannel type; matching more than one subchannel
> > > > > type is probably not useful anyway.
> > > > >
> > > > > Signed-off-by: Cornelia Huck <[email protected]>
> > > >
> > > > Reviewed-by: Sebastian Ott <[email protected]>
> > > >
> > > > Should I take that via our git tree or do you have other patches depending
> > > > on this one?
> > > >
> > >
> > > No, this patch is stand-alone; everything else is happening in user
> > > space, so taking it via your tree would be great. Thanks!
> >
> > Friendly ping (I don't see it on s390/features yet; or is this going
> > via some other path?)
> >
>
> It is there now. Just waited until latest ci results to push the branch.
> Sorry for delay.
>
Awesome, thanks!