Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20200827' into staging

- document s390x boot devices
- bug fixes

# gpg: Signature made Thu 27 Aug 2020 13:16:41 BST
# gpg:                using RSA key C3D0D66DC3624FF6A8C018CEDECF6B93C6F02FAF
# gpg:                issuer "cohuck@redhat.com"
# gpg: Good signature from "Cornelia Huck <conny@cornelia-huck.de>" [unknown]
# gpg:                 aka "Cornelia Huck <huckc@linux.vnet.ibm.com>" [full]
# gpg:                 aka "Cornelia Huck <cornelia.huck@de.ibm.com>" [full]
# gpg:                 aka "Cornelia Huck <cohuck@kernel.org>" [unknown]
# gpg:                 aka "Cornelia Huck <cohuck@redhat.com>" [unknown]
# Primary key fingerprint: C3D0 D66D C362 4FF6 A8C0  18CE DECF 6B93 C6F0 2FAF

* remotes/cohuck/tags/s390x-20200827:
  hw/vfio/ap: Plug memleak in vfio_ap_get_group()
  docs/system/s390x: Add a chapter about s390x boot devices
  virtio-ccw-input: fix description

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

4 files changed, 85 insertions, 1 deletions

diff --git a/docs/system/s390x/bootdevices.rst b/docs/system/s390x/bootdevices.rst
new file mode 100644
index 0000000000..9e591cb9dc
--- /dev/null
+++ b/docs/system/s390x/bootdevices.rst
@@ -0,0 +1,82 @@
+Boot devices on s390x
+=====================
+
+Booting with bootindex parameter
+--------------------------------
+
+For classical mainframe guests (i.e. LPAR or z/VM installations), you always
+have to explicitly specify the disk where you want to boot from (or "IPL" from,
+in s390x-speak -- IPL means "Initial Program Load"). In particular, there can
+also be only one boot device according to the architecture specification, thus
+specifying multiple boot devices is not possible (yet).
+
+So for booting an s390x guest in QEMU, you should always mark the
+device where you want to boot from with the ``bootindex`` property, for
+example::
+
+ qemu-system-s390x -drive if=none,id=dr1,file=guest.qcow2 \
+                   -device virtio-blk,drive=dr1,bootindex=1
+
+For booting from a CD-ROM ISO image (which needs to include El-Torito boot
+information in order to be bootable), it is recommended to specify a ``scsi-cd``
+device, for example like this::
+
+ qemu-system-s390x -blockdev file,node-name=c1,filename=... \
+                   -device virtio-scsi \
+                   -device scsi-cd,drive=c1,bootindex=1
+
+Note that you really have to use the ``bootindex`` property to select the
+boot device. The old-fashioned ``-boot order=...`` command of QEMU (and
+also ``-boot once=...``) is not supported on s390x.
+
+
+Booting without bootindex parameter
+-----------------------------------
+
+The QEMU guest firmware (the so-called s390-ccw bios) has also some rudimentary
+support for scanning through the available block devices. So in case you did
+not specify a boot device with the ``bootindex`` property, there is still a
+chance that it finds a bootable device on its own and starts a guest operating
+system from it. However, this scanning algorithm is still very rough and may
+be incomplete, so that it might fail to detect a bootable device in many cases.
+It is really recommended to always specify the boot device with the
+``bootindex`` property instead.
+
+This also means that you should avoid the classical short-cut commands like
+``-hda``, ``-cdrom`` or ``-drive if=virtio``, since it is not possible to
+specify the ``bootindex`` with these commands. Note that the convenience
+``-cdrom`` option even does not give you a real (virtio-scsi) CD-ROM device on
+s390x. Due to technical limitations in the QEMU code base, you will get a
+virtio-blk device with this parameter instead, which might not be the right
+device type for installing a Linux distribution via ISO image. It is
+recommended to specify a CD-ROM device via ``-device scsi-cd`` (as mentioned
+above) instead.
+
+
+Booting from a network device
+-----------------------------
+
+Beside the normal guest firmware (which is loaded from the file ``s390-ccw.img``
+in the data directory of QEMU, or via the ``-bios`` option), QEMU ships with
+a small TFTP network bootloader firmware for virtio-net-ccw devices, too. This
+firmware is loaded from a file called ``s390-netboot.img`` in the QEMU data
+directory. In case you want to load it from a different filename instead,
+you can specify it via the ``-global s390-ipl.netboot_fw=filename``
+command line option.
+
+The ``bootindex`` property is especially important for booting via the network.
+If you don't specify the the ``bootindex`` property here, the network bootloader
+firmware code won't get loaded into the guest memory so that the network boot
+will fail. For a successful network boot, try something like this::
+
+ qemu-system-s390x -netdev user,id=n1,tftp=...,bootfile=... \
+                   -device virtio-net-ccw,netdev=n1,bootindex=1
+
+The network bootloader firmware also has basic support for pxelinux.cfg-style
+configuration files. See the `PXELINUX Configuration page
+<https://wiki.syslinux.org/wiki/index.php?title=PXELINUX#Configuration>`__
+for details how to set up the configuration file on your TFTP server.
+The supported configuration file entries are ``DEFAULT``, ``LABEL``,
+``KERNEL``, ``INITRD`` and ``APPEND`` (see the `Syslinux Config file syntax
+<https://wiki.syslinux.org/wiki/index.php?title=Config>`__ for more
+information).
diff --git a/docs/system/target-s390x.rst b/docs/system/target-s390x.rst
index 644e404ef9..c636f64113 100644
--- a/docs/system/target-s390x.rst
+++ b/docs/system/target-s390x.rst
@@ -31,4 +31,5 @@ Architectural features
 ======================
 
 .. toctree::
+   s390x/bootdevices
    s390x/protvirt
diff --git a/hw/s390x/virtio-ccw-input.c b/hw/s390x/virtio-ccw-input.c
index 5601e25dee..83136fbba1 100644
--- a/hw/s390x/virtio-ccw-input.c
+++ b/hw/s390x/virtio-ccw-input.c
@@ -1,5 +1,5 @@
 /*
- * virtio ccw scsi implementation
+ * virtio ccw input implementation
  *
  * Copyright 2012, 2015 IBM Corp.
  *
diff --git a/hw/vfio/ap.c b/hw/vfio/ap.c
index b9330a8e6f..cec6fe1599 100644
--- a/hw/vfio/ap.c
+++ b/hw/vfio/ap.c
@@ -71,6 +71,7 @@ static VFIOGroup *vfio_ap_get_group(VFIOAPDevice *vapdev, Error **errp)
     if (!group_path) {
         error_setg(errp, "%s: no iommu_group found for %s: %s",
                    VFIO_AP_DEVICE_TYPE, vapdev->vdev.sysfsdev, gerror->message);
+        g_error_free(gerror);
         return NULL;
     }