sdk/emulator/qemu.git
14 years agospice: add keyboard
Gerd Hoffmann [Thu, 11 Mar 2010 14:13:28 +0000 (11:13 -0300)]
spice: add keyboard

Open keyboard channel.  Now you can type into the spice client and the
keyboard events are sent to your guest.  You'll need some other display
like vnc to actually see the guest responding to them though.

14 years agospice: core bits
Gerd Hoffmann [Thu, 11 Mar 2010 14:13:27 +0000 (11:13 -0300)]
spice: core bits

Add -spice command line switch.  Has support setting passwd and port for
now.  With this patch applied the spice client can successfully connect
to qemu.  You can't do anything useful yet though.

14 years agoadd spice into the configure file
Gerd Hoffmann [Wed, 24 Mar 2010 09:26:51 +0000 (10:26 +0100)]
add spice into the configure file

14 years agoconfigure: add logging
Gerd Hoffmann [Fri, 23 Apr 2010 11:44:10 +0000 (13:44 +0200)]
configure: add logging

Write compile commands and messages to config.log.
Useful for debugging configure.

14 years agoadd pflib: PixelFormat conversion library.
Gerd Hoffmann [Mon, 14 Jun 2010 07:54:27 +0000 (09:54 +0200)]
add pflib: PixelFormat conversion library.

14 years agoUse machine_init() to register virtfs config options.
Gerd Hoffmann [Wed, 25 Aug 2010 10:19:49 +0000 (12:19 +0200)]
Use machine_init() to register virtfs config options.

14 years agoUse display types for local display only.
Gerd Hoffmann [Wed, 25 Aug 2010 08:51:06 +0000 (10:51 +0200)]
Use display types for local display only.

This patch drops DT_VNC.  The display types are only used to select
select the local display (i.e. curses, sdl, coca, ...).  Remote
displays (for now only vnc, spice will follow) can be enabled
independently.

14 years agoMerge remote branch 'mst/for_anthony' into staging
Anthony Liguori [Mon, 20 Sep 2010 18:22:20 +0000 (13:22 -0500)]
Merge remote branch 'mst/for_anthony' into staging

14 years agoe1000: Pad short frames to minimum size (60 bytes)
Stefan Hajnoczi [Sat, 18 Sep 2010 20:43:45 +0000 (21:43 +0100)]
e1000: Pad short frames to minimum size (60 bytes)

The OpenIndiana (Solaris) e1000g driver drops frames that are too long
or too short.  It expects to receive frames of at least the Ethernet
minimum size.  ARP requests in particular are small and will be dropped
if they are not padded appropriately, preventing a Solaris VM from
becoming visible on the network.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agomingw: use ASLR, no-SEH and DEP if available
Blue Swirl [Sun, 19 Sep 2010 08:36:34 +0000 (08:36 +0000)]
mingw: use ASLR, no-SEH and DEP if available

If the linker supports the flags --dynamicbase, --no-seh,
or --nxcompat, use them.

Tested on Windows Vista: Process Explorer reports that ASLR and DEP
are in use. No effect seen on Wine or Windows XP.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agocris: Fix watchdog resets
Edgar E. Iglesias [Sat, 18 Sep 2010 22:30:25 +0000 (00:30 +0200)]
cris: Fix watchdog resets

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agocris: Fix TLB exec bit protection
Edgar E. Iglesias [Sat, 18 Sep 2010 10:34:59 +0000 (12:34 +0200)]
cris: Fix TLB exec bit protection

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agoUse gcc warning flag -Wnested-externs
Blue Swirl [Sat, 18 Sep 2010 07:02:16 +0000 (07:02 +0000)]
Use gcc warning flag -Wnested-externs

If the compiler supports the warning flag -Wnested-externs, use it.

Avoid the only warning by moving the declaration of xml_builtin to a
more proper place.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoUse gcc warning flag -Wempty-body
Blue Swirl [Sat, 18 Sep 2010 07:01:48 +0000 (07:01 +0000)]
Use gcc warning flag -Wempty-body

If the compiler supports the warning flag -Wempty-body, use it.

Adjust the code to avoid the warnings.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoUse a few more gcc warning flags
Blue Swirl [Sat, 18 Sep 2010 07:01:05 +0000 (07:01 +0000)]
Use a few more gcc warning flags

If the compiler supports the following warning flags, use them:

-Wformat-security -Wformat-y2k -Winit-self -Wignored-qualifiers
-Wmissing-include-dirs

Currently, these flags don't produce any warnings.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoUse gcc warning flag -Wtype-limits
Blue Swirl [Sat, 18 Sep 2010 05:53:15 +0000 (05:53 +0000)]
Use gcc warning flag -Wtype-limits

If the compiler supports the warning flag -Wtype-limits, use it.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agopxa2xx: fix SSSR TFN logic
Blue Swirl [Sat, 18 Sep 2010 05:53:15 +0000 (05:53 +0000)]
pxa2xx: fix SSSR TFN logic

Fix SSSR TFN logic: TX FIFO is never filled, so it is always in
underrun condition if SSP is enabled.

This also avoids a gcc warning with -Wtype-limits.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoMIPS: fix yield handling
Blue Swirl [Sat, 18 Sep 2010 05:53:15 +0000 (05:53 +0000)]
MIPS: fix yield handling

The parameter for yield should be handled as a signed integer
for the comparisons to have any effect.

This also avoids a gcc warning with -Wtype-limits.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoPPC: Suppress gcc warnings with -Wtype-limits
Blue Swirl [Sat, 18 Sep 2010 05:53:15 +0000 (05:53 +0000)]
PPC: Suppress gcc warnings with -Wtype-limits

The hack added by c5b76b381081680633e2e0a91216507430409fb2 was not
enough to avoid warnings with gcc flag -Wtype-limits. Add a new macro
to fix both problems.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoblkdebug: fix enum comparison
Blue Swirl [Sat, 18 Sep 2010 05:53:15 +0000 (05:53 +0000)]
blkdebug: fix enum comparison

The signedness of enum types depend on the compiler implementation.
Therefore the check for negative values may or may not be meaningful.

Fix by explicitly casting to a signed integer.

Since the values are also checked earlier against event_names
table, this is an internal error. Change the 'if' to 'assert'.

This also avoids a warning with GCC flag -Wtype-limits.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agopxa2xx: remove useless checks
Blue Swirl [Sat, 18 Sep 2010 05:53:15 +0000 (05:53 +0000)]
pxa2xx: remove useless checks

Remove checks which were made useless by r5849,
8da3ff180974732fc4272cb4433fef85c1822961.

This also avoids a warning with GCC flag -Wtype-limits.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoUse range_covers_byte
Blue Swirl [Sat, 18 Sep 2010 05:53:14 +0000 (05:53 +0000)]
Use range_covers_byte

Use range_covers_byte() instead of comparisons.

This avoids some warnings with GCC flag -Wtype-limits.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoIntroduce range.h
Blue Swirl [Sat, 18 Sep 2010 05:53:14 +0000 (05:53 +0000)]
Introduce range.h

Extract range functions from pci.h. These will be used by later patches
by non-PCI devices. Adjust current users.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agolinux-user: improve flatload error checking
Blue Swirl [Sat, 18 Sep 2010 05:53:14 +0000 (05:53 +0000)]
linux-user: improve flatload error checking

Because of the use of unsigned type, possible errors during
load were ignored.

Fix by using a signed type.

This also avoids a warning with GCC flag -Wtype-limits.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agolinux-user: fix types in a comparison
Blue Swirl [Sat, 18 Sep 2010 05:53:14 +0000 (05:53 +0000)]
linux-user: fix types in a comparison

-1ul is unsigned long, which does not necessarily match abi_ulong
type.

Fix by using abi_long instead.

This also avoids a warning with GCC flag -Wtype-limits.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agolinux-user: fix socklen_t comparisons
Blue Swirl [Sat, 18 Sep 2010 05:53:14 +0000 (05:53 +0000)]
linux-user: fix socklen_t comparisons

On many systems, socklen_t is defined as unsigned. This means that
checks for negative values are not meaningful.

Fix by explicitly casting to a signed integer.

This also avoids some warnings with GCC flag -Wtype-limits.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoCheck for errors during BIOS or kernel load
Blue Swirl [Sat, 18 Sep 2010 05:53:14 +0000 (05:53 +0000)]
Check for errors during BIOS or kernel load

Because of the use of unsigned types, possible errors during
BIOS or kernel load were ignored.

Fix by using a signed type.

This also avoids some warnings with GCC flag -Wtype-limits.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoMake ARP replies at least 64 bytes long
Hervé Poussineau [Wed, 15 Sep 2010 20:33:26 +0000 (22:33 +0200)]
Make ARP replies at least 64 bytes long

IEEE 802.3 standard requires Ethernet frames to be at least 64 bytes long.
If it is not the case, they will be considered as runt frames, and may be ignored by netcard and/or OS

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agoAccept packets with TTL=1
Hervé Poussineau [Mon, 13 Sep 2010 21:01:30 +0000 (23:01 +0200)]
Accept packets with TTL=1

Packets with TTL=1 may be directed to local network (DHCP/DNS servers for example), so don't discard them
This is required by old versions of NetBSD which send DHCP DISCOVER packets with TTL=1

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agocris: Avoid spurios hw_abort on recursive bus faults
Edgar E. Iglesias [Thu, 16 Sep 2010 13:40:27 +0000 (15:40 +0200)]
cris: Avoid spurios hw_abort on recursive bus faults

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agoPPC: Redesign interrupt trigger path
Alexander Graf [Fri, 10 Sep 2010 15:08:34 +0000 (15:08 +0000)]
PPC: Redesign interrupt trigger path

According to the Book3S spec, the interrupt context starts with an MSR
value that is rather simple. If we leave out the HV case, it's almost
always 0.

To reflect this, let's redesign the way that MSR value gets calculated.
Using this, we also squash the bug where MSR_POW can slip through into
the interrupt handler MSR.

Reported-by: Thomas Monjalon <thomas.monjalon@openwide.fr>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agoPPC: Enable hint bits for lwarx/ldarx
Alexander Graf [Fri, 10 Sep 2010 15:08:33 +0000 (15:08 +0000)]
PPC: Enable hint bits for lwarx/ldarx

The lwarx and ldarx instructions have a bit to give some hint to the
CPU which is safe to ignore. We currently refuse to accept any instruction
with that bit set, as it used to be declared MBZ.

Let's remove the reserved bit and make the instruction work as expected.
This fixes Linux boot for ppc64.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Andreas Färber <andreas.faerber@web.de>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agoserial: Update parameters after load
Michal Novotny [Wed, 15 Sep 2010 13:35:53 +0000 (15:35 +0200)]
serial: Update parameters after load

This is the patch to update serial port parameters after guest is
already loaded.

Signed-off-by: Michal Novotny <minovotn@redhat.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agovhost: fix infinite loop on error path
Michael S. Tsirkin [Tue, 14 Sep 2010 11:48:17 +0000 (13:48 +0200)]
vhost: fix infinite loop on error path

file.index is unsigned, hence 'while (--file.index >= 0)'
will loop > forever. Change to while (file.index-- > 0).

Reported-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
14 years agoRemove wrong semicolon in macro definition
Stefan Weil [Mon, 13 Sep 2010 19:21:57 +0000 (21:21 +0200)]
Remove wrong semicolon in macro definition

Macros normally should not end with a semicolon,
otherwise their usage results in two statements
where only one statement was expected.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agoserial: Wrap debug prints around a DPRINTF macro
Michal Novotny [Mon, 13 Sep 2010 12:32:32 +0000 (14:32 +0200)]
serial: Wrap debug prints around a DPRINTF macro

Signed-off-by: Michal Novotny <minovotn@redhat.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agoESP: fix ESP DMA access when DMA is not enabled
Blue Swirl [Sat, 11 Sep 2010 16:38:33 +0000 (16:38 +0000)]
ESP: fix ESP DMA access when DMA is not enabled

Sending ESP a command caused it to trigger DMA immediately
even if DMA was not enabled at the DMA controller.

Add a signal from DMA controller to ESP to tell ESP about changes in
DMA enable bit. Also use the correct function for setting up GPIO outputs.

This fixes NetBSD 1.6.1 through 3.0 boot.

Thanks to Artyom Tarasenko for extensive debugging of the problem.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agopowerpc: Avoid TLB related log spamming
Edgar E. Iglesias [Sat, 11 Sep 2010 12:29:07 +0000 (14:29 +0200)]
powerpc: Avoid TLB related log spamming

Invalid TLB entries are normal and should not spam the log.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
14 years agotrace: fix a regex portability problem
Blue Swirl [Sat, 11 Sep 2010 10:43:39 +0000 (10:43 +0000)]
trace: fix a regex portability problem

The /bin/sh in Milax has problems with the regex:
Error: invalid trace backend
Please choose a supported trace backend.

Fix it by escaping ')' like the regexes with '('.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agotrace: fix a typo
Blue Swirl [Fri, 10 Sep 2010 20:54:56 +0000 (23:54 +0300)]
trace: fix a typo

There is no qemu_valloc() but qemu_vmalloc().

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoHACKING: add rules for printf-like functions
Blue Swirl [Fri, 10 Sep 2010 18:53:11 +0000 (18:53 +0000)]
HACKING: add rules for printf-like functions

Add rules for printf-like functions, based on libvirt HACKING.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoHACKING: add string management rules
Blue Swirl [Fri, 10 Sep 2010 18:53:05 +0000 (18:53 +0000)]
HACKING: add string management rules

Add string management rules, somewhat like libvirt HACKING.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoHACKING: add memory management rules
Blue Swirl [Fri, 10 Sep 2010 18:52:52 +0000 (18:52 +0000)]
HACKING: add memory management rules

Add memory management rules, somewhat like libvirt HACKING.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoHACKING: add C type rules
Blue Swirl [Fri, 10 Sep 2010 18:47:31 +0000 (18:47 +0000)]
HACKING: add C type rules

Add C type rules, adapted from libvirt HACKING. Also include
a description of special QEMU scalar types.

Move typedef rule from CODING_STYLE rule 3 to HACKING rule 6
where it belongs.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoHACKING: add preprocessor rules
Blue Swirl [Fri, 10 Sep 2010 18:46:00 +0000 (18:46 +0000)]
HACKING: add preprocessor rules

Add a new file, HACKING, in order to collect recurring
issues with submitted patches.

Start with preprocessor rules, adapted from libvirt HACKING.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agovmware_vga: Replace the few tab-indents with spaces (clean-up).
Andrzej Zaborowski [Fri, 10 Sep 2010 00:30:04 +0000 (02:30 +0200)]
vmware_vga: Replace the few tab-indents with spaces (clean-up).

Use 4 spaces instead of 2, too.

14 years agovmware_vga: Add checks to deal with non-atomic fifo writes.
Andrzej Zaborowski [Fri, 10 Sep 2010 00:23:31 +0000 (02:23 +0200)]
vmware_vga: Add checks to deal with non-atomic fifo writes.

Janne Huttunen noticed that the FIFO end pointer is updated by the
guest after writing each word to the FIFO, at least the X.org driver
which is open does this.  This means that there's no way for the
host to know if the guest is in the middle a write operation.  Qemu
thus needs to read the beginning of the command up to when it's able
to tell how many words are expected for the given command.  It will
abort reading and rewind the FIFO if there aren't enough words yet,
this should be relatively rare but it is suspected to have been the
cause of the occasional FIFO overrun that killed the display.

14 years agohw/omap: Fix default setup for OMAP UART devices
Stefan Weil [Sun, 8 Aug 2010 12:09:26 +0000 (14:09 +0200)]
hw/omap: Fix default setup for OMAP UART devices

Character devices created by qemu_chr_open don't
allow duplicate device names, so naming all
UART devices "null" no longer works.

Running "qemu-system-arm -M n800" (and some other machines)
results in this error message:

qemu-system-arm: Duplicate ID 'null' for chardev
Can't create serial device, empty char device

This is fixed by setting a default label "uart1",
"uart2" or "uart3".

Cc: Andrzej Zaborowski <andrew.zaborowski@intel.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
14 years agotrace: Trace entry point of balloon request handler
Prerna Saxena [Wed, 11 Aug 2010 11:46:03 +0000 (17:16 +0530)]
trace: Trace entry point of balloon request handler

Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Trace port IO
Prerna Saxena [Wed, 11 Aug 2010 11:45:11 +0000 (17:15 +0530)]
trace: Trace port IO

Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Trace virtqueue operations
Stefan Hajnoczi [Mon, 24 May 2010 12:19:21 +0000 (13:19 +0100)]
trace: Trace virtqueue operations

This patch adds trace events for virtqueue operations including
adding/removing buffers, notifying the guest, and receiving a notify
from the guest.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Trace virtio-blk, multiwrite, and paio_submit
Stefan Hajnoczi [Sat, 22 May 2010 17:15:08 +0000 (18:15 +0100)]
trace: Trace virtio-blk, multiwrite, and paio_submit

This patch adds trace events that make it possible to observe
virtio-blk.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Trace qemu_malloc() and qemu_vmalloc()
Stefan Hajnoczi [Sat, 22 May 2010 17:09:25 +0000 (18:09 +0100)]
trace: Trace qemu_malloc() and qemu_vmalloc()

It is often useful to instrument memory management functions in order to
find leaks or performance problems.  This patch adds trace events for
the memory allocation primitives.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Add user documentation
Stefan Hajnoczi [Tue, 22 Jun 2010 14:07:09 +0000 (15:07 +0100)]
trace: Add user documentation

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Add LTTng Userspace Tracer backend
Stefan Hajnoczi [Sat, 22 May 2010 20:11:33 +0000 (21:11 +0100)]
trace: Add LTTng Userspace Tracer backend

This patch adds LTTng Userspace Tracer (UST) backend support.  The UST
system requires no kernel support but libust and liburcu must be
installed.

$ ./configure --trace-backend ust
$ make

Start the UST daemon:
$ ustd &

List available tracepoints and enable some:
$ ustctl --list-markers $(pgrep qemu)
[...]
{PID: 5458, channel/marker: ust/paio_submit, state: 0, fmt: "acb %p
opaque %p sector_num %lu nb_sectors %lu type %lu" 0x4b32ba}
$ ustctl --enable-marker "ust/paio_submit" $(pgrep qemu)

Run the trace:
$ ustctl --create-trace $(pgrep qemu)
$ ustctl --start-trace $(pgrep qemu)
[...]
$ ustctl --stop-trace $(pgrep qemu)
$ ustctl --destroy-trace $(pgrep qemu)

Trace results can be viewed using lttv-gui.

More information about UST:
http://lttng.org/ust

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
trace: Check for LTTng Userspace Tracer headers

When using the 'ust' backend, check if the relevant headers are
available at host.

Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Add trace file name command-line option
Prerna Saxena [Mon, 9 Aug 2010 10:48:32 +0000 (11:48 +0100)]
trace: Add trace file name command-line option

This patch adds an optional command line switch '-trace' to specify the
filename to write traces to, when qemu starts.
Eg, If compiled with the 'simple' trace backend,
[temp@system]$ qemu -trace FILENAME IMAGE
Allows the binary traces to be written to FILENAME instead of the option
set at config-time.

Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Add trace-file command to open/close/flush trace file
Stefan Hajnoczi [Tue, 13 Jul 2010 08:26:33 +0000 (09:26 +0100)]
trace: Add trace-file command to open/close/flush trace file

This patch adds the trace-file command:

  trace-file [on|off|flush]

  Open, close, or flush the trace file.  If no argument is given,
  the status of the trace file is displayed.

The trace file is turned on by default but is only written out when the
trace buffer becomes full.  The flush operation can be used to force
write out at any time.

Turning off the trace file does not change the state of trace events;
tracing will continue to the trace buffer.  When the trace file is off,
use "info trace" to display the contents of the trace buffer in memory.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
This commit also contains the trace-file sub-command from the following
commit:

commit 5ce8d1a957afae2c52ad748944ce72848ccf57bd
Author: Prerna Saxena <prerna@linux.vnet.ibm.com>
Date:   Wed Aug 4 16:23:54 2010 +0530

    trace: Add options to specify trace file name at startup and runtime

    This patch adds an optional command line switch '-trace' to specify the
    filename to write traces to, when qemu starts.
    Eg, If compiled with the 'simple' trace backend,
    [temp@system]$ qemu -trace FILENAME IMAGE
    Allows the binary traces to be written to FILENAME instead of the option
    set at config-time.

    Also, this adds monitor sub-command 'set' to trace-file commands to
    dynamically change trace log file at runtime.
    Eg,
    (qemu)trace-file set FILENAME
    This allows one to set trace outputs to FILENAME from the default
    specified at startup.

Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Specify trace file name
Prerna Saxena [Tue, 13 Jul 2010 08:26:32 +0000 (09:26 +0100)]
trace: Specify trace file name

Allow users to specify a file for trace-outputs at configuration.
Also, allow trace files to be annotated by <pid> so each qemu instance has
unique traces.

The trace file name can be passed as a config option:
--trace-file=/path/to/file
(Default: trace )
At runtime, the pid of the qemu process is appended to the filename so
that mutiple qemu instances do not have overlapping logs.

Eg : trace-1234 for qemu launched with pid 1234.

I have yet to test this on windows. getpid() is used at many places
in code(including vnc.c), so I'm hoping this would be okay too.

Edited-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Support disabled events in trace-events
Stefan Hajnoczi [Mon, 24 May 2010 10:32:09 +0000 (11:32 +0100)]
trace: Support disabled events in trace-events

Sometimes it is useful to disable a trace event.  Removing the event
from trace-events is not enough since source code will call the
trace_*() function for the event.

This patch makes it easy to build without specific trace events by
marking them disabled in trace-events:

disable multiwrite_cb(void *mcb, int ret) "mcb %p ret %d"

This builds without the multiwrite_cb trace event.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
trace: Allow bulk enabling/disabling of trace events at compile time

For 'simple' trace backend, allow bulk enabling/disabling of trace
events at compile time.  Trace events that are preceded by 'disable'
keyword are compiled in, but turned off by default. These can
individually be turned on using the monitor.  All other trace events are
enabled by default.

TODO :
This could be enhanced when the trace-event namespace is partitioned into a
group and an ID within that group. In such a case, marking a group as enabled
would automatically enable all trace-events listed under it.

Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Support for dynamically enabling/disabling trace events
Prerna Saxena [Thu, 24 Jun 2010 11:34:53 +0000 (17:04 +0530)]
trace: Support for dynamically enabling/disabling trace events

This patch adds support for dynamically enabling/disabling of trace events.
This is done by internally maintaining each trace event's state, and
permitting logging of data from a trace event only if it is in an
'active' state.

Monitor commands added :
1) info trace-events  : to view all available trace events and
  their state.
2) trace-event NAME on|off  : to enable/disable data logging from a
  given trace event.
  Eg, trace-event paio_submit off
   disables logging of data when
paio_submit is hit.

By default, all trace-events are disabled. One can enable desired trace-events
via the monitor.

Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
trace: Monitor command 'info trace'

Monitor command 'info trace' to display contents of trace buffer

Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
trace: Remove monitor.h dependency from simpletrace

User-mode targets don't have a monitor so the simple trace backend
currently does not build on those targets.  This patch abstracts the
monitor printing interface so there is no direct coupling between
simpletrace and the monitor.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Add simple built-in tracing backend
Stefan Hajnoczi [Sat, 22 May 2010 18:24:51 +0000 (19:24 +0100)]
trace: Add simple built-in tracing backend

This patch adds a simple tracer which produces binary trace files.  To
try out the simple backend:

$ ./configure --trace-backend=simple
$ make

After running QEMU you can pretty-print the trace:

$ ./simpletrace.py trace-events trace.log

The output of simpletrace.py looks like this:

  qemu_realloc 0.699 ptr=0x24363f0 size=0x3 newptr=0x24363f0
  qemu_free 0.768 ptr=0x24363f0
  ^           ^---- timestamp delta (us)
  |____ trace event name

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
trace: Make trace record fields 64-bit

Explicitly use 64-bit fields in trace records so that timestamps and
magic numbers work for 32-bit host builds.

Includes fixes from Prerna Saxena <prerna@linux.vnet.ibm.com>.

Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agotrace: Add trace-events file for declaring trace events
Stefan Hajnoczi [Sat, 22 May 2010 16:52:39 +0000 (17:52 +0100)]
trace: Add trace-events file for declaring trace events

This patch introduces the trace-events file where trace events can be
declared like so:

qemu_malloc(size_t size) "size %zu"
qemu_free(void *ptr) "ptr %p"

These trace event declarations are processed by a new tool called
tracetool to generate code for the trace events.  Trace event
declarations are independent of the backend tracing system (LTTng User
Space Tracing, ftrace markers, DTrace).

The default "nop" backend generates empty trace event functions.
Therefore trace events are disabled by default.

The trace-events file serves two purposes:

1. Adding trace events is easy.  It is not necessary to understand the
   details of a backend tracing system.  The trace-events file is a
   single location where trace events can be declared without code
   duplication.

2. QEMU is not tightly coupled to one particular backend tracing system.
   In order to support tracing across QEMU host platforms and to
   anticipate new backend tracing systems that are currently maturing,
   it is important to be flexible and not tied to one system.

This commit includes fixes from Prerna Saxena
<prerna@linux.vnet.ibm.com> and Blue Swirl <blauwirbel@gmail.com>.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
14 years agomicroblaze: Add support for fcmp.un
Edgar E. Iglesias [Thu, 9 Sep 2010 20:05:48 +0000 (22:05 +0200)]
microblaze: Add support for fcmp.un

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@petalogix.com>
14 years agoelf: Calculate symbol size if needed
Stefan Weil [Mon, 9 Aug 2010 14:43:53 +0000 (16:43 +0200)]
elf: Calculate symbol size if needed

Symbols with a size of 0 are unusable for the disassembler.

Example:

While running an arm linux kernel, no symbolic names are
used in qemu.log when the cpu is executing an assembler function.

Assume that the size of such symbols is the difference to the
next symbol value.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agoFix OpenBSD build warning
Blue Swirl [Thu, 9 Sep 2010 19:13:04 +0000 (19:13 +0000)]
Fix OpenBSD build warning

Fix this warning:
  CC    savevm.o
/src/qemu/savevm.c: In function `do_savevm':
/src/qemu/savevm.c:1900: warning: passing arg 1 of `localtime_r' from incompatible pointer type

It looks like on OpenBSD the type of tv_sec in struct timeval is still
'long' instead of time_t as in most other OS. Fix by adding a cast.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
14 years agomicroblaze: User-mode emulation of hw-excp signals
Edgar E. Iglesias [Thu, 9 Sep 2010 08:24:01 +0000 (10:24 +0200)]
microblaze: User-mode emulation of hw-excp signals

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@petalogix.com>
14 years agomicroblaze: Add basic FPU emulation
Edgar E. Iglesias [Thu, 9 Sep 2010 08:20:17 +0000 (10:20 +0200)]
microblaze: Add basic FPU emulation

Missing:
* fcmp.un insn
* Denormalized exceptions
* Exception model is not accurate

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@petalogix.com>
14 years agomicroblaze: Add definitions for FSR reg fields
Edgar E. Iglesias [Thu, 9 Sep 2010 07:58:35 +0000 (09:58 +0200)]
microblaze: Add definitions for FSR reg fields

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@petalogix.com>
14 years agoRevert "Make default invocation of block drivers safer (v3)"
Anthony Liguori [Wed, 8 Sep 2010 22:09:15 +0000 (17:09 -0500)]
Revert "Make default invocation of block drivers safer (v3)"

This reverts commit 79368c81bf8cf93864d7afc88b81b05d8f0a2c90.

Conflicts:

block.c

I haven't been able to come up with a solution yet for the corruption caused by
unaligned requests from the IDE disk so revert until a solution can be written.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoRevert "PPC: Qdev'ify e500 pci"
Anthony Liguori [Wed, 8 Sep 2010 19:55:19 +0000 (14:55 -0500)]
Revert "PPC: Qdev'ify e500 pci"

This reverts commit 13b7fdeffa68e3382231a70308593ae6a75d96c3.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoRevert "PPC: Make e500 pci byte swap config data"
Anthony Liguori [Wed, 8 Sep 2010 19:54:58 +0000 (14:54 -0500)]
Revert "PPC: Make e500 pci byte swap config data"

This reverts commit cfb207e643d94e3e96d456b1df14c5e36f6aa9e5.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
14 years agoMerge remote branch 'agraf/ppc-next' into staging
Anthony Liguori [Wed, 8 Sep 2010 19:29:13 +0000 (14:29 -0500)]
Merge remote branch 'agraf/ppc-next' into staging

14 years agoMerge remote branch 'kwolf/for-anthony' into staging
Anthony Liguori [Wed, 8 Sep 2010 19:26:57 +0000 (14:26 -0500)]
Merge remote branch 'kwolf/for-anthony' into staging

14 years agoMerge remote branch 'mst/for_anthony' into staging
Anthony Liguori [Wed, 8 Sep 2010 19:26:14 +0000 (14:26 -0500)]
Merge remote branch 'mst/for_anthony' into staging

14 years agovirtio-9p: Change handling of flags in open() path for 9P2000.L
Sripathi Kodi [Fri, 20 Aug 2010 10:47:47 +0000 (16:17 +0530)]
virtio-9p: Change handling of flags in open() path for 9P2000.L

This patch applies on top of 9P2000.L patches that we have on the list.
I took a look at how 9P server is handling open() flags in 9P2000.L path.
I think we can do away with the valid_flags() function and simplify the
code. The reasoning is as follows:

O_NOCTTY: (If the file is a terminal, don't make it the controlling
terminal of the process even though the process does not have a controlling
terminal) By the time the control reaches 9P client it is clear that what
we have is not a terminal device. Hence it does not matter what we do with
this flag. In any case 9P server can filter this flag out before making the
syscall.

O_NONBLOCK: (Don't block if i) Can't read/write to the file ii) Can't get
locks) This has an impact on FIFOs, but also on file locks. Hence we can
pass it down to the system call.

O_ASYNC: From the manpage:

   O_ASYNC
          Enable signal-driven I/O: generate a signal (SIGIO by default,  but
          this  can be changed via fcntl(2)) when input or output becomes pos-
          sible on this file descriptor.  This feature is only available  for
          terminals,  pseudo-terminals,  sockets,  and (since Linux 2.6) pipes
          and FIFOs.  See fcntl(2) for further details.

Again, this does not make any impact on regular files handled by 9P. Also,
we don't want 9P server to receive SIGIO. Hence I think 9P server can
filter this flag out before making the syscall.

O_CLOEXEC: This flag makes sense only on the client. If guest user space
sets this flag the guest VFS will take care of calling close() on the fd if
an exec() happens. Hence 9P client need not be bothered with this flag.
Also I think QEMU will not do an exec, but if it does, it makes sense to
close these fds. Hence we can pass this flag down to the syscall.

O_CREAT: Since we are in open() path it means we have confirmed that the file
exists. Hence there is no need to pass O_CREAT flag down to the system. In fact
on some versions of glibc this causes problems, because we pass O_CREAT flag,
but don't have permission bits. Hence we can just mask this flag out.

So in summary:

Mask out:
O_NOCTTY
O_ASYNC
O_CREAT

Pass-through:
O_NONBLOCK
O_CLOEXEC

Signed-off-by: Sripathi Kodi <sripathik@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years ago[virtio-9p] This patch implements TLERROR/RLERROR on the qemu 9P server.
Arun R Bharadwaj [Wed, 28 Jul 2010 08:40:22 +0000 (14:10 +0530)]
[virtio-9p] This patch implements TLERROR/RLERROR on the qemu 9P server.

Signed-off-by: Arun R Bharadwaj <arun@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years ago[virtio-9p] Remove all instances of unnecessary dotu variable.
Arun R Bharadwaj [Wed, 28 Jul 2010 08:25:05 +0000 (13:55 +0530)]
[virtio-9p] Remove all instances of unnecessary dotu variable.

Signed-off-by: Arun R Bharadwaj <arun@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years agovirtio-9p: Add support for removing xattr
Aneesh Kumar K.V [Thu, 26 Aug 2010 05:45:23 +0000 (11:15 +0530)]
virtio-9p: Add support for removing xattr

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
14 years agovirtio-9p: Fix the memset usage
Aneesh Kumar K.V [Thu, 2 Sep 2010 05:39:07 +0000 (11:09 +0530)]
virtio-9p: Fix the memset usage

The arguments are wrong. Use qemu_mallocz directly

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
14 years agovirtio-9p: Use lchown which won't follow symlink
Aneesh Kumar K.V [Thu, 2 Sep 2010 05:39:07 +0000 (11:09 +0530)]
virtio-9p: Use lchown which won't follow symlink

We should always use functions which don't follow
symlink on the server

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
14 years agovirtio-9p: Add SM_NONE security model
Aneesh Kumar K.V [Thu, 2 Sep 2010 05:39:07 +0000 (11:09 +0530)]
virtio-9p: Add SM_NONE security model

This is equivalent to SM_PASSTHROUGH security model.
The only exception is, failure of privilige operation like chown
are ignored. This makes a passthrough like security model usable
for people who runs kvm as non root

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
14 years agovirtio-9p: Hide user.virtfs xattr in case of mapped security.
Aneesh Kumar K.V [Thu, 2 Sep 2010 05:39:07 +0000 (11:09 +0530)]
virtio-9p: Hide user.virtfs xattr in case of mapped security.

With mapped security mode we use "user.virtfs" namespace is used
to store the virtFs related attributes. So hide it from user.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
14 years agovirtio-9p: Implement TXATTRCREATE
Aneesh Kumar K.V [Thu, 2 Sep 2010 05:39:07 +0000 (11:09 +0530)]
virtio-9p: Implement TXATTRCREATE

TXATTRCREATE:  Prepare a fid for setting xattr value on a file system object.

 size[4] TXATTRCREATE tag[2] fid[4] name[s] attr_size[8] flags[4]
 size[4] RXATTRWALK tag[2]

txattrcreate gets a fid pointing to xattr. This fid can later be
used to get set the xattr value.

flag value is derived from set Linux setxattr. The manpage says
"The flags parameter can be used to refine the semantics of the operation.
XATTR_CREATE specifies a pure create, which fails if the named attribute
exists already. XATTR_REPLACE specifies a pure replace operation, which
fails if the named attribute does not already exist. By default (no flags),
the extended attribute will be created if need be, or will simply replace
the value if the attribute exists."

The actual setxattr operation happens when the fid is clunked. At that point
the written byte count and the attr_size specified in TXATTRCREATE should be
same otherwise an error will be returned.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
14 years agovirtio-9p: Implement TXATTRWALK
Aneesh Kumar K.V [Thu, 2 Sep 2010 05:39:06 +0000 (11:09 +0530)]
virtio-9p: Implement TXATTRWALK

TXATTRWALK: Descend a ATTR namespace

 size[4] TXATTRWALK tag[2] fid[4] newfid[4] name[s]
 size[4] RXATTRWALK tag[2] size[8]

txattrwalk gets a fid pointing to xattr. This fid can later be
used to get read the xattr value. If name is NULL the fid returned
can be used to get the list of extended attribute associated to
the file system object.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
14 years agovirtio-9p: Add fidtype so that we can do type specific operation
Aneesh Kumar K.V [Thu, 2 Sep 2010 05:39:06 +0000 (11:09 +0530)]
virtio-9p: Add fidtype so that we can do type specific operation

We want to add type specific operation during read/write

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
14 years ago[virtio-9p] qemu: virtio-9p: Implement LOPEN
M. Mohan Kumar [Tue, 22 Jun 2010 14:17:04 +0000 (19:47 +0530)]
[virtio-9p] qemu: virtio-9p: Implement LOPEN

Implement 9p2000.L version of open(LOPEN) interface in qemu 9p server.

For LOPEN, no need to convert the flags to and from 9p mode to VFS mode.

Synopsis:

    size[4] Tlopen tag[2] fid[4] mode[4]

    size[4] Rlopen tag[2] qid[13] iounit[4]

Current qemu 9p server does not support following flags:
    O_NOCTTY, O_NONBLOCK, O_ASYNC & O_CLOEXEC

[Fix mode format - jvrao@linux.vnet.ibm.com]

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years agorename - change name of file or directory
M. Mohan Kumar [Tue, 22 Jun 2010 06:59:41 +0000 (12:29 +0530)]
rename - change name of file or directory

size[4] Trename tag[2] fid[4] newdirfid[4] name[s]
size[4] Rrename tag[2]

Implement the 2000.L rename operation. A new function
v9fs_complete_rename is introduced that acts as a common entry point
for 2000.L rename operation and 2000.U rename opearation (via wstat).
As part of this change the field 'nname' (used only for rename) is
removed from the structure V9fsWstatState. Instead a new structure
V9fsRenameState is used for rename operations both by 2000.U and 2000.L
code paths. Both 2000.U and 2000.L rename code paths construct the
V9fsRenameState structure and passes that to v9fs_complete_rename
function.

Changes from previous version:
 Use qemu_mallocz to initialize
 Use strcpy,strcat functions instead of memcpy
 Changed the variable name to newdirfid
 Introduced post rename function
 Error checking
 Removed nname field from V9fsWstatState

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years agoqemu: virtio-9p: Implement TMKDIR
M. Mohan Kumar [Tue, 22 Jun 2010 06:55:22 +0000 (12:25 +0530)]
qemu: virtio-9p: Implement TMKDIR

Synopsis

    size[4] Tmkdir tag[2] fid[4] name[s] mode[4] gid[4]

    size[4] Rmkdir tag[2] qid[13]

Description

    mkdir asks the file server to create a directory with given name,
    mode and gid. The qid for the new directory is returned with
    the mkdir reply message.

Note: 72 is selected as the opcode for TMKDIR from the reserved list.

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
[jvrao@linux.vnet.ibm.com: Fix perm handling when creating directory]

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years agoqemu: virtio-9p: Implement TMKNOD
M. Mohan Kumar [Tue, 22 Jun 2010 06:54:09 +0000 (12:24 +0530)]
qemu: virtio-9p: Implement TMKNOD

Implement TMKNOD as part of 2000.L Work

Synopsis

    size[4] Tmknod tag[2] fid[4] name[s] mode[4] major[4] minor[4] gid[4]

    size[4] Rmknod tag[2] qid[13]

Description

    mknod asks the file server to create a device node with given device
    type, mode and gid. The qid for the new device node is returned with
    the mknod reply message.

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years ago[virtio-9p] This patch implements TLCREATE for 9p2000.L protocol.
Venkateswararao Jujjuri (JV) [Fri, 18 Jun 2010 01:27:24 +0000 (18:27 -0700)]
[virtio-9p]  This patch implements TLCREATE for 9p2000.L protocol.

SYNOPSIS

    size[4] Tlcreate tag[2] fid[4] name[s] flags[4] mode[4] gid[4]

    size[4] Rlcreate tag[2] qid[13] iounit[4]

DESCRIPTION

The Tlreate request asks the file server to create a new regular file with the
name supplied, in the directory (dir) represented by fid.
The mode argument specifies the permissions to use. New file is created with
the uid if the fid and with supplied gid.

The flags argument represent Linux access mode flags with which the caller
is requesting to open the file with. Protocol allows all the Linux access
modes but it is upto the server to allow/disallow any of these acess modes.
If the server doesn't support any of the access mode, it is expected to
return error.

To start with we will not restricit/limit any Linux flags on this server.
If needed, We can start restricting as we move forward with various use cases.

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years ago[virtio-9p] Define and implement TSYMLINK for 9P2000.L
Venkateswararao Jujjuri (JV) [Wed, 9 Jun 2010 21:02:08 +0000 (14:02 -0700)]
[virtio-9p] Define and implement TSYMLINK for 9P2000.L

This patch implements creating a symlink for TSYMLINK request
and responds with RSYMLINK. In the case of error, we return RERROR.

SYNOPSIS

    size[4] Tsymlink tag[2] fid[4] name[s] symtgt[s] gid[4]

    size[4] Rsymlink tag[2] qid[13]

    DESCRIPTION

    Create a symbolic link named 'name' pointing to 'symtgt'.
    gid represents the effective group id of the caller.
    The  permissions of a symbolic link are irrelevant hence it is omitted
    from the protocol.

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years ago[virtio-9p] Implement TLINK for 9P2000.L
Venkateswararao Jujjuri (JV) [Wed, 9 Jun 2010 18:21:15 +0000 (11:21 -0700)]
[virtio-9p] Implement TLINK for 9P2000.L

Create a Hardlink.

SYNOPSIS

size[4] Tlink tag[2] dfid[4] oldfid[4] newpath[s]

size[4] Rlink tag[2]

DESCRIPTION

Create a link 'newpath' in directory pointed by dfid linking to oldfid path.

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years agovirtio-9p: Implement server side of setattr for 9P2000.L protocol.
Sripathi Kodi [Thu, 17 Jun 2010 12:48:47 +0000 (18:18 +0530)]
virtio-9p: Implement server side of setattr for 9P2000.L protocol.

SYNOPSIS

      size[4] Tsetattr tag[2] attr[n]

      size[4] Rsetattr tag[2]

   DESCRIPTION

      The setattr command changes some of the file status information.
      attr resembles the iattr structure used in Linux kernel. It
      specifies which status parameter is to be changed and to what
      value. It is laid out as follows:

         valid[4]
            specifies which status information is to be changed. Possible
            values are:
            ATTR_MODE       (1 << 0)
            ATTR_UID        (1 << 1)
            ATTR_GID        (1 << 2)
            ATTR_SIZE       (1 << 3)
            ATTR_ATIME      (1 << 4)
            ATTR_MTIME      (1 << 5)
            ATTR_CTIME      (1 << 5)
            ATTR_ATIME_SET  (1 << 7)
            ATTR_MTIME_SET  (1 << 8)

            The last two bits represent whether the time information
            is being sent by the client's user space. In the absense
            of these bits the server always uses server's time.

         mode[4]
            File permission bits

         uid[4]
            Owner id of file

         gid[4]
            Group id of the file

         size[8]
            File size

         atime_sec[8]
            Time of last file access, seconds

         atime_nsec[8]
            Time of last file access, nanoseconds

         mtime_sec[8]
            Time of last file modification, seconds

         mtime_nsec[8]
            Time of last file modification, nanoseconds

Explanation of the patches:
--------------------------

*) The kernel just copies relevent contents of iattr structure to p9_iattr_dotl
   structure and passes it down to the client. The only check it has is calling
   inode_change_ok()
*) The p9_iattr_dotl structure does not have ctime and ia_file parameters because
   I don't think these are needed in our case. The client user space can request
   updating just ctime by calling chown(fd, -1, -1). This is handled on server
   side without a need for putting ctime on the wire.
*) The server currently supports changing mode, time, ownership and size of the
   file.
*) 9P RFC says "Either all the changes in wstat request happen, or none of them
   does: if the request succeeds, all changes were made; if it fails, none were."
   I have not done anything to implement this specifically because I don't see
   a reason.

[jvrao@linux.vnet.ibm.com: Parts of code for handling chown(-1,-1)

Signed-off-by: Sripathi Kodi <sripathik@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years ago[virtio-9p] Make v9fs_do_utimensat accept timespec structures instead of v9stat.
Sripathi Kodi [Wed, 9 Jun 2010 13:44:38 +0000 (19:14 +0530)]
[virtio-9p] Make v9fs_do_utimensat accept timespec structures instead of v9stat.

Currently v9fs_do_utimensat takes a V9fsStat argument and builds
timespec structures. It sets tv_nsec values to 0 by default. Instead
of this it should take struct timespec[2] and pass it down to the
system directly. This will make it more generic and useful
elsewhere.

Signed-off-by: Sripathi Kodi <sripathik@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years agovirtio-9p: Do not reset atime
M. Mohan Kumar [Wed, 9 Jun 2010 13:44:38 +0000 (19:14 +0530)]
virtio-9p: Do not reset atime

    Current code resets file's atime to 0 when there is a change in mtime.
    This results in resetting the atime to "1970-01-01 05:30:00". For
    example, truncate -s 0 filename results in changing the mtime to the
    truncate time, but resets the atime to "1970-01-01 05:30:00". utime
    system call does not have any provision to set only mtime or atime. So
    change v9fs_wstat_post_chmod function to use utimensat function to change
    the atime and mtime fields. If tv_nsec field is set to the special value
    "UTIME_OMIT", corresponding file time stamp is not updated.

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years agovirtio-9p: getattr server implementation for 9P2000.L protocol.
Sripathi Kodi [Tue, 20 Jul 2010 06:14:41 +0000 (11:44 +0530)]
virtio-9p: getattr server implementation for 9P2000.L protocol.

           SYNOPSIS

              size[4] Tgetattr tag[2] fid[4] request_mask[8]

              size[4] Rgetattr tag[2] lstat[n]

           DESCRIPTION

              The getattr transaction inquires about the file identified by fid.
              request_mask is a bit mask that specifies which fields of the
              stat structure is the client interested in.

              The reply will contain a machine-independent directory entry,
              laid out as follows:

                 st_result_mask[8]
                    Bit mask that indicates which fields in the stat structure
                    have been populated by the server

                 qid.type[1]
                    the type of the file (directory, etc.), represented as a bit
                    vector corresponding to the high 8 bits of the file's mode
                    word.

                 qid.vers[4]
                    version number for given path

                 qid.path[8]
                    the file server's unique identification for the file

                 st_mode[4]
                    Permission and flags

                 st_uid[4]
                    User id of owner

                 st_gid[4]
                    Group ID of owner

                 st_nlink[8]
                    Number of hard links

                 st_rdev[8]
                    Device ID (if special file)

                 st_size[8]
                    Size, in bytes

                 st_blksize[8]
                    Block size for file system IO

                 st_blocks[8]
                    Number of file system blocks allocated

                 st_atime_sec[8]
                    Time of last access, seconds

                 st_atime_nsec[8]
                    Time of last access, nanoseconds

                 st_mtime_sec[8]
                    Time of last modification, seconds

                 st_mtime_nsec[8]
                    Time of last modification, nanoseconds

                 st_ctime_sec[8]
                    Time of last status change, seconds

                 st_ctime_nsec[8]
                    Time of last status change, nanoseconds

                 st_btime_sec[8]
                    Time of creation (birth) of file, seconds

                 st_btime_nsec[8]
                    Time of creation (birth) of file, nanoseconds

                 st_gen[8]
                    Inode generation

                 st_data_version[8]
                    Data version number

              request_mask and result_mask bit masks contain the following bits
                 #define P9_STATS_MODE          0x00000001ULL
                 #define P9_STATS_NLINK         0x00000002ULL
                 #define P9_STATS_UID           0x00000004ULL
                 #define P9_STATS_GID           0x00000008ULL
                 #define P9_STATS_RDEV          0x00000010ULL
                 #define P9_STATS_ATIME         0x00000020ULL
                 #define P9_STATS_MTIME         0x00000040ULL
                 #define P9_STATS_CTIME         0x00000080ULL
                 #define P9_STATS_INO           0x00000100ULL
                 #define P9_STATS_SIZE          0x00000200ULL
                 #define P9_STATS_BLOCKS        0x00000400ULL

                 #define P9_STATS_BTIME         0x00000800ULL
                 #define P9_STATS_GEN           0x00001000ULL
                 #define P9_STATS_DATA_VERSION  0x00002000ULL

                 #define P9_STATS_BASIC         0x000007ffULL
                 #define P9_STATS_ALL           0x00003fffULL

        This patch implements the client side of getattr implementation for 9P2000.L.
        It introduces a new structure p9_stat_dotl for getting Linux stat information
        along with QID. The data layout is similar to stat structure in Linux user
        space with the following major differences:

        inode (st_ino) is not part of data. Instead qid is.

        device (st_dev) is not part of data because this doesn't make sense on the
        client.

        All time variables are 64 bit wide on the wire. The kernel seems to use
        32 bit variables for these variables. However, some of the architectures
        have used 64 bit variables and glibc exposes 64 bit variables to user
        space on some architectures. Hence to be on the safer side we have made
        these 64 bit in the protocol. Refer to the comments in
        include/asm-generic/stat.h

        There are some additional fields: st_btime_sec, st_btime_nsec, st_gen,
        st_data_version apart from the bitmask, st_result_mask. The bit mask
        is filled by the server to indicate which stat fields have been
        populated by the server. Currently there is no clean way for the
        server to obtain these additional fields, so it sends back just the
        basic fields.

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Signed-off-by: Sripathi Kodi <sripathik@in.ibm.com>
14 years agovirtio-9p: Compute iounit based on host filesystem block size
M. Mohan Kumar [Wed, 9 Jun 2010 13:44:28 +0000 (19:14 +0530)]
virtio-9p: Compute iounit based on host filesystem block size

Compute iounit based on the host filesystem block size and pass it to
client with open/create response. Also return iounit as statfs's f_bsize
for optimal block size transfers.

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Reviewd-by: Sripathi Kodi <sripathik@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years ago[V4] virtio-9p: readdir implementation for 9p2000.L
Sripathi Kodi [Wed, 9 Jun 2010 09:27:57 +0000 (14:57 +0530)]
[V4] virtio-9p: readdir implementation for 9p2000.L

This patch implements the server part of readdir() implementation for
9p2000.L

    SYNOPSIS

    size[4] Treaddir tag[2] fid[4] offset[8] count[4]
    size[4] Rreaddir tag[2] count[4] data[count]

    DESCRIPTION

    The readdir request asks the server to read the directory specified by 'fid'
    at an offset specified by 'offset' and return as many dirent structures as
    possible that fit into count bytes. Each dirent structure is laid out as
    follows.

            qid.type[1]
              the type of the file (directory, etc.), represented as a bit
              vector corresponding to the high 8 bits of the file's mode
              word.

            qid.vers[4]
              version number for given path

            qid.path[8]
              the file server's unique identification for the file

            offset[8]
              offset into the next dirent.

            type[1]
              type of this directory entry.

            name[256]
              name of this directory entry.

Signed-off-by: Sripathi Kodi <sripathik@in.ibm.com>
Reviewed-by: M. Mohan Kumar <mohan@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years agovirtio-9p: Return correct error from v9fs_remove
Sripathi Kodi [Wed, 9 Jun 2010 09:03:22 +0000 (14:33 +0530)]
virtio-9p: Return correct error from v9fs_remove

Signed-off-by: Sripathi Kodi <sripathik@in.ibm.com>
In v9fs_remove_post_remove() we currently ignore the error returned by
the previous call to remove() and return an error only if freeing the
fid fails. However, the client expects to see the error from remove().
Currently the client falsely thinks that the remove call has always
succeeded. For example, doing rmdir on a non-empty directory does
not return ENOTEMPTY.

With this patch we ignore the error from free_fid(). The client cannot
use this error value anyway.

Signed-off-by: Sripathi Kodi <sripathik@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
14 years agoqemu: virtio-9p: Implement statfs support in server
M. Mohan Kumar [Mon, 10 May 2010 06:41:03 +0000 (12:11 +0530)]
qemu: virtio-9p: Implement statfs support in server

Implement statfs support in qemu server based on Sripathi's
initial statfs patch.

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Signed-off-by: Sripathi Kodi <sripathik@in.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>