Re: [Autotest] [PATCH] KVM test: Add PCI device assignment support

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Jan 18, 2010 at 02:49:14PM -0200, Lucas Meneghel Rodrigues wrote:
> Ok, after some consideration I have applied the patch:
> 
> http://autotest.kernel.org/changeset/4144
> 
> Thank you for your work on this test!

I should thank you for improving this test. :-)

> 
> On Wed, Jan 13, 2010 at 12:19 AM, Yolkfull Chow <yzhou@xxxxxxxxxx> wrote:
> > On Tue, Jan 12, 2010 at 04:28:13PM -0200, Lucas Meneghel Rodrigues wrote:
> >> Add support to PCI device assignment on the kvm test. It supports
> >> both SR-IOV virtual functions and physical NIC card device
> >> assignment.
> >>
> >> Single Root I/O Virtualization (SR-IOV) allows a single PCI device to
> >> be shared amongst multiple virtual machines while retaining the
> >> performance benefit of assigning a PCI device to a virtual machine.
> >> A common example is where a single SR-IOV capable NIC - with perhaps
> >> only a single physical network port - might be shared with multiple
> >> virtual machines by assigning a virtual function to each VM.
> >>
> >> SR-IOV support is implemented in the kernel. The core implementation
> >> is contained in the PCI subsystem, but there must also be driver support
> >> for both the Physical Function (PF) and Virtual Function (VF) devices.
> >> With an SR-IOV capable device one can allocate VFs from a PF. The VFs
> >> surface as PCI devices which are backed on the physical PCI device by
> >> resources (queues, and register sets).
> >>
> >> Device support:
> >>
> >> In 2.6.30, the Intel® 82576 Gigabit Ethernet Controller is the only
> >> SR-IOV capable device supported. The igb driver has PF support and the
> >> igbvf has VF support.
> >>
> >> In 2.6.31 the Neterion® X3100™ is supported as well. This device uses
> >> the same vxge driver for the PF as well as the VFs.
> >>
> >> In order to configure the test:
> >>
> >>   * For SR-IOV virtual functions passthrough, we could specify the
> >>     module parameter 'max_vfs' in config file.
> >>   * For physical NIC card pass through, we should specify the device
> >>     name(s).
> >
> > Looks good for me.
> > Lucas, thank you so much for improving this patch. :-)
> >
> >>
> >> Signed-off-by: Yolkfull Chow <yzhou@xxxxxxxxxx>
> >> Signed-off-by: Lucas Meneghel Rodrigues <lmr@xxxxxxxxxx>
> >> ---
> >>  client/tests/kvm/kvm_utils.py          |  284 ++++++++++++++++++++++++++++++++
> >>  client/tests/kvm/kvm_vm.py             |   60 +++++++
> >>  client/tests/kvm/tests_base.cfg.sample |   22 +++-
> >>  3 files changed, 365 insertions(+), 1 deletions(-)
> >>
> >> diff --git a/client/tests/kvm/kvm_utils.py b/client/tests/kvm/kvm_utils.py
> >> index 2bbbe22..a2d9607 100644
> >> --- a/client/tests/kvm/kvm_utils.py
> >> +++ b/client/tests/kvm/kvm_utils.py
> >> @@ -924,3 +924,287 @@ def create_report(report_dir, results_dir):
> >>      reporter = os.path.join(report_dir, 'html_report.py')
> >>      html_file = os.path.join(results_dir, 'results.html')
> >>      os.system('%s -r %s -f %s -R' % (reporter, results_dir, html_file))
> >> +
> >> +
> >> +def get_full_pci_id(pci_id):
> >> +    """
> >> +    Get full PCI ID of pci_id.
> >> +
> >> +    @param pci_id: PCI ID of a device.
> >> +    """
> >> +    cmd = "lspci -D | awk '/%s/ {print $1}'" % pci_id
> >> +    status, full_id = commands.getstatusoutput(cmd)
> >> +    if status != 0:
> >> +        return None
> >> +    return full_id
> >> +
> >> +
> >> +def get_vendor_from_pci_id(pci_id):
> >> +    """
> >> +    Check out the device vendor ID according to pci_id.
> >> +
> >> +    @param pci_id: PCI ID of a device.
> >> +    """
> >> +    cmd = "lspci -n | awk '/%s/ {print $3}'" % pci_id
> >> +    return re.sub(":", " ", commands.getoutput(cmd))
> >> +
> >> +
> >> +class PciAssignable(object):
> >> +    """
> >> +    Request PCI assignable devices on host. It will check whether to request
> >> +    PF (physical Functions) or VF (Virtual Functions).
> >> +    """
> >> +    def __init__(self, type="nic_vf", driver=None, driver_option=None,
> >> +                 names=None, devices_requested=None):
> >> +        """
> >> +        Initialize parameter 'type' which could be:
> >> +        nic_vf: Virtual Functions
> >> +        nic_pf: Physical Function (actual hardware)
> >> +        mixed:  Both includes VFs and PFs
> >> +
> >> +        If pass through Physical NIC cards, we need to specify which devices
> >> +        to be assigned, e.g. 'eth1 eth2'.
> >> +
> >> +        If pass through Virtual Functions, we need to specify how many vfs
> >> +        are going to be assigned, e.g. passthrough_count = 8 and max_vfs in
> >> +        config file.
> >> +
> >> +        @param type: PCI device type.
> >> +        @param driver: Kernel module for the PCI assignable device.
> >> +        @param driver_option: Module option to specify the maximum number of
> >> +                VFs (eg 'max_vfs=7')
> >> +        @param names: Physical NIC cards correspondent network interfaces,
> >> +                e.g.'eth1 eth2 ...'
> >> +        @param devices_requested: Number of devices being requested.
> >> +        """
> >> +        self.type = type
> >> +        self.driver = driver
> >> +        self.driver_option = driver_option
> >> +        if names:
> >> +            self.name_list = names.split()
> >> +        if devices_requested:
> >> +            self.devices_requested = int(devices_requested)
> >> +        else:
> >> +            self.devices_requested = None
> >> +
> >> +
> >> +    def _get_pf_pci_id(self, name, search_str):
> >> +        """
> >> +        Get the PF PCI ID according to name.
> >> +
> >> +        @param name: Name of the PCI device.
> >> +        @param search_str: Search string to be used on lspci.
> >> +        """
> >> +        cmd = "ethtool -i %s | awk '/bus-info/ {print $2}'" % name
> >> +        s, pci_id = commands.getstatusoutput(cmd)
> >> +        if not (s or "Cannot get driver information" in pci_id):
> >> +            return pci_id[5:]
> >> +        cmd = "lspci | awk '/%s/ {print $1}'" % search_str
> >> +        pci_ids = [id for id in commands.getoutput(cmd).splitlines()]
> >> +        nic_id = int(re.search('[0-9]+', name).group(0))
> >> +        if (len(pci_ids) - 1) < nic_id:
> >> +            return None
> >> +        return pci_ids[nic_id]
> >> +
> >> +
> >> +    def _release_dev(self, pci_id):
> >> +        """
> >> +        Release a single PCI device.
> >> +
> >> +        @param pci_id: PCI ID of a given PCI device.
> >> +        """
> >> +        base_dir = "/sys/bus/pci"
> >> +        full_id = get_full_pci_id(pci_id)
> >> +        vendor_id = get_vendor_from_pci_id(pci_id)
> >> +        drv_path = os.path.join(base_dir, "devices/%s/driver" % full_id)
> >> +        if 'pci-stub' in os.readlink(drv_path):
> >> +            cmd = "echo '%s' > %s/new_id" % (vendor_id, drv_path)
> >> +            if os.system(cmd):
> >> +                return False
> >> +
> >> +            stub_path = os.path.join(base_dir, "drivers/pci-stub")
> >> +            cmd = "echo '%s' > %s/unbind" % (full_id, stub_path)
> >> +            if os.system(cmd):
> >> +                return False
> >> +
> >> +            driver = self.dev_drivers[pci_id]
> >> +            cmd = "echo '%s' > %s/bind" % (full_id, driver)
> >> +            if os.system(cmd):
> >> +                return False
> >> +
> >> +        return True
> >> +
> >> +
> >> +    def get_vf_devs(self):
> >> +        """
> >> +        Catch all VFs PCI IDs.
> >> +
> >> +        @return: List with all PCI IDs for the Virtual Functions avaliable
> >> +        """
> >> +        if not self.sr_iov_setup():
> >> +            return []
> >> +
> >> +        cmd = "lspci | awk '/Virtual Function/ {print $1}'"
> >> +        return commands.getoutput(cmd).split()
> >> +
> >> +
> >> +    def get_pf_devs(self):
> >> +        """
> >> +        Catch all PFs PCI IDs.
> >> +
> >> +        @return: List with all PCI IDs for the physical hardware requested
> >> +        """
> >> +        pf_ids = []
> >> +        for name in self.name_list:
> >> +            pf_id = self._get_pf_pci_id(name, "Ethernet")
> >> +            if not pf_id:
> >> +                continue
> >> +            pf_ids.append(pf_id)
> >> +        return pf_ids
> >> +
> >> +
> >> +    def get_devs(self, count):
> >> +        """
> >> +        Check out all devices' PCI IDs according to their name.
> >> +
> >> +        @param count: count number of PCI devices needed for pass through
> >> +        @return: a list of all devices' PCI IDs
> >> +        """
> >> +        if self.type == "nic_vf":
> >> +            vf_ids = self.get_vf_devs()
> >> +        elif self.type == "nic_pf":
> >> +            vf_ids = self.get_pf_devs()
> >> +        elif self.type == "mixed":
> >> +            vf_ids = self.get_vf_devs()
> >> +            vf_ids.extend(self.get_pf_devs())
> >> +        return vf_ids[0:count]
> >> +
> >> +
> >> +    def get_vfs_count(self):
> >> +        """
> >> +        Get VFs count number according to lspci.
> >> +        """
> >> +        cmd = "lspci | grep 'Virtual Function' | wc -l"
> >> +        # For each VF we'll see 2 prints of 'Virtual Function', so let's
> >> +        # divide the result per 2
> >> +        return int(commands.getoutput(cmd)) / 2
> >> +
> >> +
> >> +    def check_vfs_count(self):
> >> +        """
> >> +        Check VFs count number according to the parameter driver_options.
> >> +        """
> >> +        return (self.get_vfs_count == self.devices_requested)
> >> +
> >> +
> >> +    def is_binded_to_stub(self, full_id):
> >> +        """
> >> +        Verify whether the device with full_id is already binded to pci-stub.
> >> +
> >> +        @param full_id: Full ID for the given PCI device
> >> +        """
> >> +        base_dir = "/sys/bus/pci"
> >> +        stub_path = os.path.join(base_dir, "drivers/pci-stub")
> >> +        if os.path.exists(os.path.join(stub_path, full_id)):
> >> +            return True
> >> +        return False
> >> +
> >> +
> >> +    def sr_iov_setup(self):
> >> +        """
> >> +        Ensure the PCI device is working in sr_iov mode.
> >> +
> >> +        Check if the PCI hardware device drive is loaded with the appropriate,
> >> +        parameters (number of VFs), and if it's not, perform setup.
> >> +
> >> +        @return: True, if the setup was completed successfuly, False otherwise.
> >> +        """
> >> +        re_probe = False
> >> +        s, o = commands.getstatusoutput('lsmod | grep %s' % self.driver)
> >> +        if s:
> >> +            re_probe = True
> >> +        elif not self.check_vfs_count():
> >> +            os.system("modprobe -r %s" % self.driver)
> >> +            re_probe = True
> >> +
> >> +        # Re-probe driver with proper number of VFs
> >> +        if re_probe:
> >> +            cmd = "modprobe %s %s" % (self.driver, self.driver_option)
> >> +            s, o = commands.getstatusoutput(cmd)
> >> +            if s:
> >> +                return False
> >> +            if not self.check_vfs_count():
> >> +                return False
> >> +            return True
> >> +
> >> +
> >> +    def request_devs(self):
> >> +        """
> >> +        Implement setup process: unbind the PCI device and then bind it
> >> +        to the pci-stub driver.
> >> +
> >> +        @return: a list of successfully requested devices' PCI IDs.
> >> +        """
> >> +        base_dir = "/sys/bus/pci"
> >> +        stub_path = os.path.join(base_dir, "drivers/pci-stub")
> >> +
> >> +        self.pci_ids = self.get_devs(self.devices_requested)
> >> +        logging.debug("The following pci_ids were found: %s" % self.pci_ids)
> >> +        requested_pci_ids = []
> >> +        self.dev_drivers = {}
> >> +
> >> +        # Setup all devices specified for assignment to guest
> >> +        for pci_id in self.pci_ids:
> >> +            full_id = get_full_pci_id(pci_id)
> >> +            if not full_id:
> >> +                continue
> >> +            drv_path = os.path.join(base_dir, "devices/%s/driver" % full_id)
> >> +            dev_prev_driver= os.path.realpath(os.path.join(drv_path,
> >> +                                              os.readlink(drv_path)))
> >> +            self.dev_drivers[pci_id] = dev_prev_driver
> >> +
> >> +            # Judge whether the device driver has been binded to stub
> >> +            if not self.is_binded_to_stub(full_id):
> >> +                logging.debug("Binding device %s to stub" % full_id)
> >> +                vendor_id = get_vendor_from_pci_id(pci_id)
> >> +                stub_new_id = os.path.join(stub_path, 'new_id')
> >> +                unbind_dev = os.path.join(drv_path, 'unbind')
> >> +                stub_bind = os.path.join(stub_path, 'bind')
> >> +
> >> +                info_write_to_files = [(vendor_id, stub_new_id),
> >> +                                       (full_id, unbind_dev),
> >> +                                       (full_id, stub_bind)]
> >> +
> >> +                for content, file in info_write_to_files:
> >> +                    try:
> >> +                        utils.open_write_close(content, file)
> >> +                    except IOError:
> >> +                        logging.debug("Failed to write %s to file %s" %
> >> +                                      (content, file))
> >> +                        continue
> >> +
> >> +                if not self.is_binded_to_stub(full_id):
> >> +                    logging.error("Binding device %s to stub failed" %
> >> +                                  pci_id)
> >> +                    continue
> >> +            else:
> >> +                logging.debug("Device %s already binded to stub" % pci_id)
> >> +            requested_pci_ids.append(pci_id)
> >> +        self.pci_ids = requested_pci_ids
> >> +        return self.pci_ids
> >> +
> >> +
> >> +    def release_devs(self):
> >> +        """
> >> +        Release all PCI devices currently assigned to VMs back to the
> >> +        virtualization host.
> >> +        """
> >> +        try:
> >> +            for pci_id in self.dev_drivers:
> >> +                if not self._release_dev(pci_id):
> >> +                    logging.error("Failed to release device %s to host" %
> >> +                                  pci_id)
> >> +                else:
> >> +                    logging.info("Released device %s successfully" % pci_id)
> >> +        except:
> >> +            return
> >> diff --git a/client/tests/kvm/kvm_vm.py b/client/tests/kvm/kvm_vm.py
> >> index 7229b79..78a16ac 100755
> >> --- a/client/tests/kvm/kvm_vm.py
> >> +++ b/client/tests/kvm/kvm_vm.py
> >> @@ -119,6 +119,7 @@ class VM:
> >>          self.params = params
> >>          self.root_dir = root_dir
> >>          self.address_cache = address_cache
> >> +        self.pci_assignable = None
> >>
> >>          # Find available monitor filename
> >>          while True:
> >> @@ -304,6 +305,12 @@ class VM:
> >>          elif params.get("uuid"):
> >>              qemu_cmd += " -uuid %s" % params.get("uuid")
> >>
> >> +        # If the PCI assignment step went OK, add each one of the PCI assigned
> >> +        # devices to the qemu command line.
> >> +        if self.pci_assignable:
> >> +            for pci_id in self.pa_pci_ids:
> >> +                qemu_cmd += " -pcidevice host=%s" % pci_id
> >> +
> >>          return qemu_cmd
> >>
> >>
> >> @@ -392,6 +399,50 @@ class VM:
> >>                  self.uuid = f.read().strip()
> >>                  f.close()
> >>
> >> +            if not params.get("pci_assignable") == "no":
> >> +                pa_type = params.get("pci_assignable")
> >> +                pa_devices_requested = params.get("devices_requested")
> >> +
> >> +                # Virtual Functions (VF) assignable devices
> >> +                if pa_type == "vf":
> >> +                    pa_driver = params.get("driver")
> >> +                    pa_driver_option = params.get("driver_option")
> >> +                    self.pci_assignable = kvm_utils.PciAssignable(type=pa_type,
> >> +                                        driver=pa_driver,
> >> +                                        driver_option=pa_driver_option,
> >> +                                        devices_requested=pa_devices_requested)
> >> +                # Physical NIC (PF) assignable devices
> >> +                elif pa_type == "pf":
> >> +                    pa_device_names = params.get("device_names")
> >> +                    self.pci_assignable = kvm_utils.PciAssignable(type=pa_type,
> >> +                                         names=pa_device_names,
> >> +                                         devices_requested=pa_devices_requested)
> >> +                # Working with both VF and PF
> >> +                elif pa_type == "mixed":
> >> +                    pa_device_names = params.get("device_names")
> >> +                    pa_driver = params.get("driver")
> >> +                    pa_driver_option = params.get("driver_option")
> >> +                    self.pci_assignable = kvm_utils.PciAssignable(type=pa_type,
> >> +                                        driver=pa_driver,
> >> +                                        driver_option=pa_driver_option,
> >> +                                        names=pa_device_names,
> >> +                                        devices_requested=pa_devices_requested)
> >> +
> >> +                self.pa_pci_ids = self.pci_assignable.request_devs()
> >> +
> >> +                if self.pa_pci_ids:
> >> +                    logging.debug("Successfuly assigned devices: %s" %
> >> +                                  self.pa_pci_ids)
> >> +                else:
> >> +                    logging.error("No PCI assignable devices were assigned "
> >> +                                  "and 'pci_assignable' is defined to %s "
> >> +                                  "on your config file. Aborting VM creation." %
> >> +                                  pa_type)
> >> +                    return False
> >> +
> >> +            else:
> >> +                self.pci_assignable = None
> >> +
> >>              # Make qemu command
> >>              qemu_command = self.make_qemu_command()
> >>
> >> @@ -537,6 +588,8 @@ class VM:
> >>              # Is it already dead?
> >>              if self.is_dead():
> >>                  logging.debug("VM is already down")
> >> +                if self.pci_assignable:
> >> +                    self.pci_assignable.release_devs()
> >>                  return
> >>
> >>              logging.debug("Destroying VM with PID %d..." %
> >> @@ -557,6 +610,9 @@ class VM:
> >>                              return
> >>                      finally:
> >>                          session.close()
> >> +                        if self.pci_assignable:
> >> +                            self.pci_assignable.release_devs()
> >> +
> >>
> >>              # Try to destroy with a monitor command
> >>              logging.debug("Trying to kill VM with monitor command...")
> >> @@ -566,6 +622,8 @@ class VM:
> >>                  # Wait for the VM to be really dead
> >>                  if kvm_utils.wait_for(self.is_dead, 5, 0.5, 0.5):
> >>                      logging.debug("VM is down")
> >> +                    if self.pci_assignable:
> >> +                        self.pci_assignable.release_devs()
> >>                      return
> >>
> >>              # If the VM isn't dead yet...
> >> @@ -575,6 +633,8 @@ class VM:
> >>              # Wait for the VM to be really dead
> >>              if kvm_utils.wait_for(self.is_dead, 5, 0.5, 0.5):
> >>                  logging.debug("VM is down")
> >> +                if self.pci_assignable:
> >> +                    self.pci_assignable.release_devs()
> >>                  return
> >>
> >>              logging.error("Process %s is a zombie!" % self.process.get_pid())
> >> diff --git a/client/tests/kvm/tests_base.cfg.sample b/client/tests/kvm/tests_base.cfg.sample
> >> index b8f25f4..49b6d6b 100644
> >> --- a/client/tests/kvm/tests_base.cfg.sample
> >> +++ b/client/tests/kvm/tests_base.cfg.sample
> >> @@ -889,7 +889,27 @@ variants:
> >>          extra_params += " -mem-path /mnt/kvm_hugepage"
> >>
> >>
> >> +variants:
> >> +    - @no_pci_assignable:
> >> +        pci_assignable = no
> >> +    - pf_assignable:
> >> +        pci_assignable = pf
> >> +        device_names = eth1
> >> +    - vf_assignable:
> >> +        pci_assignable = vf
> >> +        # Driver (kernel module) that supports SR-IOV hardware.
> >> +        # As of today (30-11-2009), we have 2 drivers for this type of hardware:
> >> +        # Intel® 82576 Gigabit Ethernet Controller - igb
> >> +        # Neterion® X3100™ - vxge
> >> +        driver = igb
> >> +        # Driver option to specify the maximum number of virtual functions
> >> +        # (on vxge the option is , for example, is max_config_dev)
> >> +        # the default below is for the igb driver
> >> +        driver_option = "max_vfs=7"
> >> +        # Number of devices that are going to be requested.
> >> +        devices_requested = 7
> >> +
> >> +
> >>  image_name.* ?<= images/
> >>  cdrom.* ?<= isos/
> >>  steps ?<= steps/
> >> -
> >> --
> >> 1.6.5.2
> > _______________________________________________
> > Autotest mailing list
> > Autotest@xxxxxxxxxxxxxxx
> > http://test.kernel.org/cgi-bin/mailman/listinfo/autotest
> >
> 
> 
> 
> -- 
> Lucas
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux