Some functions were prototyped as static but the actual definition
wasn't. While this is valid (the function is static because the two
declarations don't conflict and the first one is static), this makes
sparse unhappy and cause confusion of normal people too.
Signed-off-by: Vincent Bernat <vincent@bernat.im>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch removes MEMCMP_IO from commontypes.h and fixes the one use of the
macro so it calls uuid_cmp_le() instead. The old code was comparing UUIDs
directly.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Remove the redundant MEMSET define in commontypes.h and fix everyplace that
uses it.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch removes UINTN from commontypes.h, using u64 in the one spot this
type was used.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Delete #defines that aren't used anywhere.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch deletes everything in common types that was in the else section
of a #ifdef __KERNEL__ block.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch switches all use of the S64 typedef to use the kernel's s64 type
instead.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Delete the S32 type from commontypes.h since it wasn't being used.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch switches all use of the S16 typedef to use the kernel's s16 type
instead.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Delete the S8 type from commontypes.h since it wasn't being used.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch switches all use of the U64 typedef to use the kernel's u64 type
instead.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch switches all use of the U32 typedef to use the kernel's u32 type
instead.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch switches all use of the U16 typedef to use the kernel's u16 type
instead.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch switches all use of the U8 typedef to use the kernel's u8 type
instead.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds new a new directory parahotplug to the visorchipset sysfs
directory, and two new attributes, deviceenabled, and devicedisabled, into the
new directory, to add s-Par parahotplug support.
The parahotplug interface is used to deal with SR-IOV recovery situations on
s-Par guest partitions. The command service partition will send a message to a
guest when an SR-IOV device that guest is using needs to be temporarily removed.
The message triggers a udev event that will cause a recovery script to run.
When that script has completed its work, it will write to one of the
parahotplug interfaces to send a message back to Command, indicating that the
recovery action has completed.
When a guest that is sharing an SR-IOV device is restarted, that guest will
take down the PF driver on the device, but any guests with VFs will not know
that their device needs to be reset as well. The recovery script makes it so
the device will be shut down fully and then restarted after the sharing guest
comes back up, and ensures that the timing is correct.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Remove the parahotplug proc code entirely in preparation for replacement with
sysfs entries. Since visorchipset_proc_read_writeonly() is orphaned by the
removal, it is removed as well.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Debugging information for the guest's channels was being exposed in proc.
Remove the code that creates these entries, which are no longer needed.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch cleans up the style, error handling, and string handling in the sysfs
functions recently added to visorchipset:
- Use of sscanf() was changed to type-appropriate kstrto*() functions
- error handling was simplified
- the error return value of visorchannel_write() was corrected
- switch use of driver-specific types to kernel types
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds a documentation file for all of the interfaces that were moved
to sysfs by the other patches in this set.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Remove remnant code left over from the diagdump proc entry.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Move the chipsetready proc entry to sysfs under a new directory guest. This
entry is used by Unisys application software on the guest to acknowledge
completion of specific events for integration purposes, but these
acknowledgements are not required for the guest to operate correctly.
The store function is simplified as well, to use scanf() instead of copying
the buffer and using strsep().
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The installer entry in /proc/visorchipset/installer was composed of three
separate fields as one entry. This patch removes the proc entry and associated
functions, and creates new fields with distinct entries under sysfs in the
visorchipset/install directory. The fields are:
textid: used to send the ID of a string that should be displayed on
s-Par's automatic installation progress screen. Setting this
field when not in installation mode (boottotool was set on
the previous guest boot) has no effect.
remaining_steps: used to set the value of the progress bar on the
s-Par automatic installation progress screen. This field has
no effect if not in installation mode.
error: used to send the ID of a string that should be displayed on
s-Par's automatic installation progress screen when an error
is encountered during installation. This field has no effect
if not in installation mode.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch fixes formatting and cleans up unneeded comments
Signed-off-by: Erik Arfvidson <erik.arfvidson@unisys.com>
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Move the proc entry controlling the boottotool flag to procfs. The field
appears in /sys/devices/platform/visorchipset/install/boottotool.
The boottotool flag controls s-Par behavior on the next boot of this guest.
Setting the flag will cause the guest to boot from the utility and installation
image, which will use the value in the toolaction field to determine what
operation is being requested.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Move the proc entry for controlling the toolaction field to sysfs. The field
appears in /sys/devices/platform/visorchipset/install/toolaction.
This field is used to tell s-Par which type of recovery tool action to perform
on the next guest boot-up. The meaning of the value is dependent on the type
of installation software used to commission the guest.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds enable_ints entry to virthba directory
Signed-off-by: Erik Arfvidson <erik.arfvidson@unisys.com>
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds the rqwait entry into the virthba debugfs directory
Signed-off-by: Erik Arfvidson <erik.arfvidson@unisys.com>
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds virthba debugfs directory and info entry
Signed-off-by: Erik Arfvidson <erik.arfvidson@unisys.com>
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch removes all proc entries, directories, and functions
Signed-off-by: Erik Arfvidson <erik.arfvidson@unisys.com>
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The controlvm channel is not removable from a guest after the guest starts,
so it makes no sense to constantly check for it. Move the channel address
discovery to visorchipset_init(), and remove all of the checks for the channel
address from the rest of the module, as the module will not load if the
channel pointer is not valid.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is no reason to have controlvm_get_channel_address() defined in a
separate file. Move it to visorchipset_main.c, remove the no-op functions
controlvm_init() and controlvm_deinit() from visorchipset_main.c, and remove
the controlvm.h and controlvm_direct.c files from the module.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds the virtpci debugfs directory and the info entry
inside of it.
Signed-off-by: Erik Arfvidson <erik.arfvidson@unisys.com>
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Remove the code that generates proc entries for the controlVM channel.
These entries are no longer necessary for debug.
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch removes all proc entries from virtpci
Signed-off-by: Erik Arfvidson <erik.arfvidson@unisys.com>
Signed-off-by: Benjamin Romer <benjamin.romer@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
vbuschannel_devinfo_to_string() was declared with the devinfo argument
as a pointer to __iomem space. No callers of this function need the
__iomem space, so I removed that constraint.
Same thing goes for vbuschannel_sanitize_buffer(). It doesn't need to
operate on I/O space.
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The function ultra_vbus_init_channel() is never called since
previous server code was removed. Remove the dead code.
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The init_vbus_channel() function is never called with the isServer
argument equal to TRUE, so just get rid of the argument altogether.
This also means that the code in the function that would be executed
in the server case can be removed since it is never executed.
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Change function names to conform to Linux kernel coding convention.
ULTRA_VBUS_init_channel() --> ultra_vbus_init_channel()
VBUSCHANNEL_devInfoToStringBuffer() --> vbuschannel_devinfo_to_string()
VBUSCHANNEL_sanitize_buffer() --> vbuschannel_sanitize_buffer()
VBUSCHANNEL_itoa() --> vbuschannel_itoa()
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The visorchannelstub module references the variable unisys_spar_platform
which is declared in the visorutil module. Add the proper dependency to
Kconfig to ensure that visorutil is enabled when building visorchannelstub.
This address the build error:
drivers/built-in.o: In function `channel_mod_init':
chanstub.c:(.init.text+0xbe24): undefined reference to `unisys_spar_platform'
Reported by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The unisys drivers now properly check to make sure they are running
on the s-Par platform before they will initialize. This was fixed in
commit fcd0157ece so it is safe to allow the unisys drivers to be built.
This has been tested in the same qemu environment that originally
produced the panic and the kernel now runs as expected.
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Tested-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The use of __DATE__ and __TIME__ is no longer allowed in the kernel so this
commit removes those. They were once useful when the drivers were being
built externally, but now that the drivers are in the kernel the use of the
macros is redundant since the kernel already has the same information
elsewhere.
In addition, using these macros breaks the build if using gcc 4.9.0
Reported-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The macros CONFIG_SPAR_GUEST, GUESTDRIVERBUILD, and NOAUTOVERSION are
defined in Makefiles for the Unisys SPAR driver. They are never used.
Remove the lines that define these macros.
Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is not any code using the functionality in filexfer.c so I removed it
and filexfer.h.
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fixed the usage of the following so they don't try to dereference
pointers to iomem.
CHANNEL_U64_MISMATCH
CHANNEL_U32_MISMATCH
wait_for_valid_guid
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
VISORCHANNEL_CHANGE_SERVER_STATE and VISORCHANNEL_CHANGE_CLIENT_STATE
are never used in any of the source so they have been removed.
VISORCHANNEL_CHANGE_CLIENT_STATE would have caused a broken kernel build
after commit a8d7f21d, but since it was never used the kernel continued
to build.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
virthba_queue_command() is only used inside virthba.c so declare it static.
Signed-off-by: Ken Cox <jkc@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add a blank line after declarations to meet kernel coding style.
Signed-off-by: Masaru Nomura <massa.nomura@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Vincent Bernat