linux-user: Implement new ARM 64 bit cmpxchg kernel helper
linux-user: Implement new ARM 64 bit cmpxchg kernel helper
Linux 3.1 will have a new kernel-page helper for ARM implementing
64 bit cmpxchg. Implement this helper in QEMU linux-user mode:
* Provide kernel helper emulation for 64bit cmpxchg
* Allow guest to object to guest offset to ensure it can map a page
* Populate page with kernel helper version
Cédric VINCENT [Fri, 26 Aug 2011 08:56:50 +0000 (10:56 +0200)]
linux-user: Fix initialization of the heap contents when allocating new pages
Technically the new mmapped pages are already initialized to zero
since they are anonymous, however we have to take care with the
contents that come from the remaining part of the previous page: it
may contains garbage data due to a previous heap usage (grown then
shrunken).
This patch completes commit 55f08c84. The problem could be reproduced
when emulating the build process of Perl 5.12.3 on ARMedSlack 13.37:
target-i386: Compute all flag data inside %cl != 0 test.
The (x << (cl - 1)) quantity is only used if CL != 0. Move the
computation of that quantity nearer its use.
This avoids the creation of undefined TCG operations when the
constant propagation optimization proves that CL == 0, and thus
CL-1 is outside the range [0-wordsize).
Juha Riihimäki [Wed, 31 Aug 2011 20:49:41 +0000 (21:49 +0100)]
usb-musb: Add reset function
Add a separate reset function musb_reset() to the usb-musb interface,
so that users who implement a reset function can also reset usb-musb.
Use this in tusb6010.
Signed-off-by: Juha Riihimäki <[email protected]>
[Riku Voipio: Fixes and restructuring patchset] Signed-off-by: Riku Voipio <[email protected]>
[Peter Maydell: More fixes and cleanups for upstream submission] Signed-off-by: Peter Maydell <[email protected]> Signed-off-by: Gerd Hoffmann <[email protected]>
Peter Maydell [Wed, 31 Aug 2011 20:49:40 +0000 (21:49 +0100)]
usb-musb: Take a DeviceState* in init function
Initialise usb-musb by passing it a DeviceState* and the offset of the
IRQs in its gpio array, rather than a plain pointer to an irq array.
This is simpler for callers and also allows us to pass in a valid parent
to usb_bus_new(), so the USB bus actually appears in the qdev tree.
This patch makes qemu assign a port when creating the device, not when
attaching it. For most usb devices this isn't a noticable difference
because they are in attached state all the time.
The change affects usb-host devices which live in detached state while
the real device is unplugged from the host. They have a fixed port
assigned all the time now instead of getting grabbing one on attach and
releasing it at detach, i.e. they stop floating around at the usb bus.
The change also allows to simplify usb-hub. It doesn't need the
handle_attach() callback any more to configure the downstream ports.
This can be done at device initialitation time now. The changed
initialization order (first grab upstream port, then register downstream
ports) also fixes some icky corner cases. For example it is not possible
any more to plug the hub into one of its own downstream ports.
The usb host adapters must care too. USBPort->dev being non-NULL
doesn't imply any more the device is in attached state. The host
adapters must additionally check the USBPort->dev->attached flag.
Gerd Hoffmann [Fri, 26 Aug 2011 12:13:48 +0000 (14:13 +0200)]
usb-ehci: handle siTDs
This patch adds code to do minimal siTD handling, which is basically
just following the next pointer. This is good enougth to handle the
inactive siTDs used by FreeBSD. Active siTDs are skipped too as we
don't have split transfer support in qemu, additionally a warning is
printed.
Gerd Hoffmann [Fri, 26 Aug 2011 14:27:41 +0000 (16:27 +0200)]
usb-host: endpoint table fixup
USB Devices can have up to 15 IN and 15 OUT endpoints, not 15 endpoints
total. Move from one array to two arrays (one IN, one OUT) to maintain
the endpoint state.
Gerd Hoffmann [Wed, 24 Aug 2011 11:34:17 +0000 (13:34 +0200)]
usb-host: claim port
When configured to pass through a specific host port (using hostbus and
hostport properties), try to claim the port if supported by the kernel.
That will avoid any kernel drivers binding to devices plugged into that
port. It will not stop any userspace apps (such as usb_modeswitch)
access the device via usbfs though.
Gerd Hoffmann [Wed, 24 Aug 2011 12:45:07 +0000 (14:45 +0200)]
usb-host: fix configuration tracking.
It is perfectly fine to leave the usb device in unconfigured state
(USBHostDevice->configuration == 0). Just do that and wait for the
guest to explicitly set a configuration. This is closer to what real
hardware does and it also simplifies the device initialization. There
is no need to figure how the device is configured on the host.
Gerd Hoffmann [Wed, 24 Aug 2011 11:45:06 +0000 (13:45 +0200)]
usb-host: limit open retries
Limit the number of times qemu tries to open host devices to three.
Reset error counter when the device goes away, after un-plugging and
re-plugging the device qemu will try again three times.
Gerd Hoffmann [Wed, 24 Aug 2011 08:55:40 +0000 (10:55 +0200)]
usb-host: fix halted endpoints
Two fixes for the price of one ;)
First, reinitialize the endpoint table after device reset.
This is needed anyway as the reset might have switched interfaces.
It also clears the endpoint halted state.
Second the CLEAR_HALT ioctl wants a unsigned int passed in as
argument, not uint8_t.
This gets my usb sd card reader (sandisk micromate) going.
Gerd Hoffmann [Mon, 22 Aug 2011 12:13:20 +0000 (14:13 +0200)]
usb-host: start tracing support
Add a bunch of trace points to usb-linux.c Drop a bunch of DPRINTK's in
favor of the trace points. Also cleanup error reporting a bit while being
at it.
spice: set qxl->ssd.running=true before telling spice to start, RHBZ #733993
If qxl->ssd.running=true is set after telling spice to start, the spice server
thread can call qxl_send_events while qxl->ssd.running is still false. This leads to
assert(d->ssd.running).
qxl: send interrupt after migration in case ram->int_pending != 0, RHBZ #732949
if qxl_send_events was called from spice server context, and then
migration had completed before a call to pipe_read, the target
guest qxl driver didn't get the interrupt. In addition,
qxl_send_events ignored further interrupts of the same kind, since
ram->int_pending was set. As a result, the guest driver was stacked
or very slow (when the waiting for the interrupt was with timeout).
Paolo Bonzini [Mon, 5 Sep 2011 16:11:51 +0000 (18:11 +0200)]
scsi: fix accounting of writes
Writes go through scsi_write_complete at least twice, the first time
to get some data without having actually written anything. Because
of this, the first time scsi_write_complete is called it will call
bdrv_acct_done and account a read incorrectly. Fix this by looking
at the aiocb. I am doing the same in scsi_read_complete for symmetry,
but it is only needed in the (bogus) case of bdrv_aio_readv returning
NULL.
Fam Zheng [Fri, 12 Aug 2011 15:19:33 +0000 (23:19 +0800)]
VMDK: bugfix, open Haiku vmdk image
Haiku provides a specially formed vmdk image, which let qemu abort. It a
combination of sparse header and flat data (i.e. with not l1/l2 table at
all). The fix is turn to descriptor when sparse header is zero in field
'capacity'.
Fam Zheng [Fri, 12 Aug 2011 15:19:32 +0000 (23:19 +0800)]
VMDK: creating streamOptimized subformat
Creating streamOptimized subformat. Added subformat option
'streamOptimized', to create a image with compression enabled and each
cluster with a GrainMarker.
spitz tosa: Simplify "drive is suitable for microdrive" test
We try the drive defined with -drive if=ide,index=0 (or equivalent
sugar). We use it only if (dinfo && bdrv_is_inserted(dinfo->bdrv) &&
!bdrv_is_removable(dinfo->bdrv)). This is a convoluted way to test
for "drive media can't be removed".
The only way to create such a drive with -drive if=ide is media=cdrom.
And that sets dinfo->media_cd, so just test that.
xen: Clean up pci_piix3_xen_ide_unplug()'s test for "not a CD"
pci_piix3_xen_ide_unplug() unplugs only disks, not CD-ROMs. It peeks
into the DriveInfo's BlockDriverState to distinguish between the two.
Unclean; use DriveInfo member media_cd, like xen_config_dev_blk().
savevm: Include writable devices with removable media
savevm and loadvm silently ignore block devices with removable media,
such as floppies and SD cards. Rolling back a VM to a previous
checkpoint will *not* roll back writes to block devices with removable
media.
Moreover, bdrv_is_removable() is a confused mess, and wrong in at
least one case: it considers "-drive if=xen,media=cdrom -M xenpv"
removable. It'll be cleaned up later in this series.
Read-only block devices are also ignored, but that's okay.
Fix by ignoring only read-only block devices and empty block devices.
Change (!bdrv_is_removable(bs) || bdrv_is_inserted(bs)) to just
bdrv_is_inserted(). Rationale:
The value of bdrv_is_removable(bs) matters only when
bdrv_is_inserted(bs) is false.
bdrv_is_inserted(bs) is true when bs is open (bs->drv != NULL) and
not an empty host drive (CD-ROM or floppy).
Therefore, bdrv_is_removable(bs) matters only when:
1. bs is not open
old: may call bdrv_flush(bs), which does nothing
new: won't call
2. bs is an empty host drive
old: may call bdrv_flush(bs), which calls driver method
raw_flush(), which calls fdatasync() or equivalent, which
can't do anything useful while the drive is empty
new: won't call
Result is bs->drv && !bdrv_is_read_only(bs) && bdrv_is_inserted(bs).
bdrv_is_inserted(bs) implies bs->drv. Drop the redundant test.
fdctrl_change_cb() gets called on a virtual media change via monitor.
It would be nice if host device block drivers called it on physical
media change, but they don't.
bdrv_media_changed() lets you poll for media change, but it returns
"don't know" except with block driver "host_floppy".
FDrive member media_changed gets set on device initialization and by
fdctrl_change_cb(), and cleared by fdctrl_media_changed(). Thus, it's
set on first entry to fdctrl_media_changed() since device
initialization or virtual media change.
fdctrl_media_changed() ignores media_changed unless
bdrv_media_changed() returns "don't know". If we change media via
monitor (setting media_changed), and the new media's block driver
returns 0, we lose. Fortunately, "host_floppy" always returns 1 on
first call. Brittle. Clean it up not to rely on it.
block/raw: Fix to forward method bdrv_media_changed()
Block driver "raw" forwards most methods to the underlying block
driver. However, it doesn't implement method bdrv_media_changed().
Makes bdrv_media_changed() always return -ENOTSUP.
I believe -fda /dev/fd0 gives you raw over host_floppy, and disk
change detection (fdc register 7 bit 7) is broken. Testing my theory
requires a computer museum, though.
ide: Update command code definitions as per ACS-2 Table B.2
Drop WIN_SRST, it has the same value as WIN_DEVICE_RESET.
Drop unused WIN_RESTORE, it has the same value as WIN_RECAL.
Drop codes that are not implemented and long obsolete: WIN_READ_LONG,
WIN_READ_LONG_ONCE, WIN_WRITE_LONG, WIN_WRITE_LONG_ONCE, WIN_FORMAT
(all obsolete since ATA4), WIN_ACKMEDIACHANGE, WIN_POSTBOOT,
WIN_PREBOOT (obsolete since ATA3), WIN_WRITE_SAME (obsolete since
ATA3, code reused for something else in ACS2), WIN_IDENTIFY_DMA
(obsolete since ATA4).
Drop codes that are not implemented and vendor-specific:
EXABYTE_ENABLE_NEST, DISABLE_SEAGATE.
Drop WIN_INIT, it isn't implemented, its value used to be reserved,
and is used for something else since ATA8.
CFA_IDLEIMMEDIATE isn't specific to CFATA. ACS-2 shows it as a
defined command in ATA-1, -2 and -3. Rename to WIN_IDLEIMMEDIATE2.
Mark vendor specific, retired, and obsolete codes.
For now, this just protects against programming errors like having the
same drive back multiple non-qdev devices, or untimely bdrv_delete().
Later commits will add other interesting uses.
While there, rename BlockDriverState member peer to dev, bdrv_attach()
to bdrv_attach_dev(), bdrv_detach() to bdrv_detach_dev(), and
bdrv_get_attached() to bdrv_get_attached_dev().
Kevin Wolf [Thu, 1 Sep 2011 14:16:10 +0000 (16:16 +0200)]
async: Allow nested qemu_bh_poll calls
qemu may segfault when a BH handler first deletes a BH and then (possibly
indirectly) calls a nested qemu_bh_poll(). This is because the inner instance
frees the BH and deletes it from the list that the outer one processes.
This patch deletes BHs only in the outermost qemu_bh_poll instance.
Commit 7887f620 already tried to achieve the same, but it assumed that the BH
handler would only delete its own BH. With a nested qemu_bh_poll(), this isn't
guaranteed, so that commit wasn't enough. Hope this one fixes it for real.
Kevin Wolf [Thu, 1 Sep 2011 13:02:13 +0000 (15:02 +0200)]
qcow2: Fix error cases to run depedent requests
Requests depending on a failed request would end up waiting forever. This fixes
the error path to continue dependent requests even when the request has failed.
Avi Kivity [Thu, 4 Aug 2011 12:55:30 +0000 (15:55 +0300)]
pflash_cfi01/pflash_cfi02: convert to memory API
cfi02 is annoying in that is ignores some address bits; we probably
want explicit support in the memory API for that.
In order to get the correct opaque into the MemoryRegion object, the
allocation scheme is changed so that the flash emulation code allocates
memory, instead of the caller. This clears a FIXME in the flash code.
Avi Kivity [Mon, 1 Aug 2011 08:04:39 +0000 (11:04 +0300)]
piix_pci: wrap memory update in a transaction
The code will remap all PAMs, even if just one is updated, resulting
in reduced performance. Wrap in a transaction to detect that those
other PAMs have not changed.
projects / qemu.git / log