Gleb Natapov [Wed, 8 Dec 2010 11:34:54 +0000 (13:34 +0200)]
Introduce fw_name field to DeviceInfo structure.
Add "fw_name" to DeviceInfo to use in device path building. In
contrast to "name" "fw_name" should refer to functionality device
provides instead of particular device model like "name" does.
Bernhard Kohl [Wed, 8 Dec 2010 14:59:55 +0000 (15:59 +0100)]
wdt_i6300esb: register a reset function
The device shall set its default hardware state after each reset.
This includes that the timer is stopped which is especially important
if the guest does a reboot independantly of a watchdog bite. I moved
the initialization of the state variables completely from the init
to the reset function which is called right after init during the
first boot and afterwards during each reboot.
Alexander Graf [Wed, 8 Dec 2010 11:05:49 +0000 (12:05 +0100)]
isa_mmio: Always use little endian
This patch converts the ISA MMIO bridge code to always use little endian mmio.
All bswap code that existed was only there to convert from native cpu
endianness to little endian ISA devices.
Alexander Graf [Wed, 8 Dec 2010 11:05:40 +0000 (12:05 +0100)]
pci-host: Delegate bswap to mmio layer
The only reason we have bswap versions of the pci host code is that
most pci host devices are little endian. The ppc e500 is the only
odd one here, being big endian.
So let's directly pass the endianness down to the mmio layer and not
worry about it on the pci host layer.
Alexander Graf [Wed, 8 Dec 2010 11:05:38 +0000 (12:05 +0100)]
Make simple io mem handler endian aware
As an alternative to the 3 individual handlers, there is also a simplified
io mem hook function. To be consistent, let's add an endianness parameter
there too.
Alexander Graf [Wed, 8 Dec 2010 11:05:37 +0000 (12:05 +0100)]
Add endianness as io mem parameter
As stated before, devices can be little, big or native endian. The
target endianness is not of their concern, so we need to push things
down a level.
This patch adds a parameter to cpu_register_io_memory that allows a
device to choose its endianness. For now, all devices simply choose
native endian, because that's the same behavior as before.
Alexander Graf [Wed, 8 Dec 2010 11:05:36 +0000 (12:05 +0100)]
exec: introduce endianness swapped mmio
The way we're currently modeling mmio is too simplified. We assume that
every device has the same endianness as the target CPU. In reality,
most devices are little endian (all PCI and ISA ones I'm aware of). Some
are big endian (special system devices) and a very little fraction is
target native endian (fw_cfg).
So instead of assuming every device to be native endianness, let's move
to a model where the device tells us which endianness it's in.
That way we can compile the devices only once and get rid of all the ugly
swap will be done by the underlying layer.
For the same of readability, this patch only introduces the helper framework
but doesn't allow the registering code to set its endianness yet.
Blue Swirl [Sat, 4 Dec 2010 17:37:35 +0000 (17:37 +0000)]
Fix mingw32 and OpenBSD warnings
ffsl() is not universally available, so there are these warnings
on both mingw32 and OpenBSD:
/src/qemu/hw/pcie_aer.c: In function 'pcie_aer_update_log':
/src/qemu/hw/pcie_aer.c:399: warning: implicit declaration of function 'ffsl'
Since status field in PCIEAERErr is uint32_t, we can just use ffs() instead.
Merge branch 'linux-user-for-upstream' of git://gitorious.org/qemu-maemo/qemu
* 'linux-user-for-upstream' of git://gitorious.org/qemu-maemo/qemu:
linux-user: fix mips and ppc to use UID16
update binfmt conf
linux-user: fix compiler error on nptl
ARM: linux-user: Restore iWMMXT state from ucontext on sigreturn
ARM: linux-user: Expose iWMMXT registers to signal handlers
ARM: linux-user: Restore VFP state from ucontext on sigreturn
ARM: linux-user: Expose VFP registers to signal handlers
ARM: Expose vfp_get_fpscr() and vfp_set_fpscr() to C code
ARM: linux-user: Correct size of padding in target_ucontext_v2
target-sparc: remove unused functions cpu_lock(), cpu_unlock()
ARM: enable XScale/iWMMXT in linux-user mode
linux-user: Translate getsockopt level option
linux-user: remove unnecessary local from __get_user(), __put_user()
linux-user: fix memory leaks with NPTL emulation
linux-user: mmap_reserve() not controlled by RESERVED_VA
[PATCH] target-arm: remove unused functions cpu_lock(), cpu_unlock()
Peter Maydell [Wed, 24 Nov 2010 15:20:06 +0000 (15:20 +0000)]
ARM: linux-user: Restore VFP state from ucontext on sigreturn
Restore the VFP registers from the ucontext on return from a signal
handler in linux-user mode. This means that signal handlers cannot
accidentally corrupt the interrupted code's VFP state, and allows
them to deliberately modify the state via the ucontext structure.
Peter Maydell [Wed, 24 Nov 2010 15:20:05 +0000 (15:20 +0000)]
ARM: linux-user: Expose VFP registers to signal handlers
For ARM linux-user mode signal handlers, fill in the ucontext with
VFP register contents in the same way that the kernel does. We only
do this for v2 format sigframe (2.6.12 and above); this is actually
bug-for-bug compatible with the older kernels, which don't save and
restore VFP registers either.
Peter Maydell [Wed, 24 Nov 2010 15:20:04 +0000 (15:20 +0000)]
ARM: Expose vfp_get_fpscr() and vfp_set_fpscr() to C code
Expose the vfp_get_fpscr() and vfp_set_fpscr() functions to C
code as well as generated code, so we can use them to read and
write the FPSCR when saving and restoring VFP registers across
signal handlers in linux-user mode.
Peter Maydell [Wed, 24 Nov 2010 15:20:03 +0000 (15:20 +0000)]
ARM: linux-user: Correct size of padding in target_ucontext_v2
The padding in the target_ucontext_v2 is defined by the size of
the target's sigset_t type, not the host's. (This bug only causes
problems when we start using the uc_regspace[] array to expose
VFP registers to userspace signal handlers.)
Jamie Lentin [Fri, 26 Nov 2010 13:04:08 +0000 (15:04 +0200)]
linux-user: Translate getsockopt level option
n setsockopt, the socket level options are translated to the hosts'
architecture before the real syscall is called, e.g.
TARGET_SO_TYPE -> SO_TYPE. This patch does the same with getsockopt.
Peter Maydell [Mon, 8 Nov 2010 18:13:58 +0000 (18:13 +0000)]
linux-user: remove unnecessary local from __get_user(), __put_user()
Remove an unnecessary local variable from the __get_user() and
__put_user() macros. This avoids confusing compilation failures
if the name of the local variable ('size') happens to be the
same as the variable the macro user is trying to read/write.
Nathan Froyd [Fri, 29 Oct 2010 14:48:57 +0000 (07:48 -0700)]
linux-user: fix memory leaks with NPTL emulation
Running programs that create large numbers of threads, such as this
snippet from libstdc++'s pthread7-rope.cc:
const int max_thread_count = 4;
const int max_loop_count = 10000;
...
for (int j = 0; j < max_loop_count; j++)
{
...
for (int i = 0; i < max_thread_count; i++)
pthread_create (&tid[i], NULL, thread_main, 0);
for (int i = 0; i < max_thread_count; i++)
pthread_join (tid[i], NULL);
}
in user-mode emulation will quickly run out of memory. This is caused
by a failure to free memory in do_syscall prior to thread exit:
/* TODO: Free CPU state. */
pthread_exit(NULL);
The first step in fixing this is to make all TaskStates used by QEMU
dynamically allocated. The TaskState used by the initial thread was
not, as it was allocated on main's stack. So fix that, free the
cpu_env, free the TaskState, and we're home free, right?
If we blindly free the TaskState, then, we yank the current (host)
thread's stack out from underneath it while it still has things to do,
like calling pthread_exit. That causes problems, as you might expect.
The solution adopted here is to let the C library allocate the thread's
stack (so the C library can properly clean it up at pthread_exit) and
provide a hint that we want NEW_STACK_SIZE bytes of stack.
With those two changes, we're done, right? Well, almost. You see,
we're creating all these host threads and their parent threads never
bother to check that their children are finished. There's no good place
for the parent threads to do so. Therefore, we need to create the
threads in a detached state so the parent thread doesn't have to call
pthread_join on the child to release the child's resources; the child
does so automatically.
With those three major changes, we can comfortably run programs like the
above without exhausting memory. We do need to delete 'stack' from the
TaskState structure.
linux-user: mmap_reserve() not controlled by RESERVED_VA
mmap_reserve() should be called only when RESERVED_VA is enabled.
Otherwise, unmaped virtual address space will never be reusable. This
bug will exhaust virtual address space in extreme conditions.
Anthony Liguori [Thu, 9 Sep 2010 19:51:31 +0000 (14:51 -0500)]
Use a Linux-style MAINTAINERS file
I make no claims that this is accurate or exhaustive but I think it's a
reasonable place to start.
As the file mentions, the purpose of this file is to give contributors
information about who they can go to with questions about a particular piece of
code or who they can ask for review.
If you sign up for a piece of code and indicate that it's Maintained or
Supported, please be prepared to be responsive to questions about that
subsystem.
Kevin Wolf [Fri, 26 Nov 2010 15:36:16 +0000 (16:36 +0100)]
ide: Reset current_addr after stopping DMA
Whenever SSBM is reset in the command register all state information is lost.
Restarting DMA means that current_addr must be reset to the base address of the
PRD table. The OS is not required to change the base address register before
starting a DMA operation, it can reuse the value it wrote for an earlier
request.
Paul Brook [Sat, 27 Nov 2010 11:23:34 +0000 (11:23 +0000)]
Split out common pcnet code
The core pcnet emulation code is used by both the PCI "pcnet" device
and the SPARC "lance" device. Split the common code frm the PCI code so
that that can be configures independantly.
Hannes Reinecke [Wed, 24 Nov 2010 11:15:59 +0000 (12:15 +0100)]
scsi: Move sense handling into the driver
The current sense handling in scsi-bus is only used by the
scsi-disk driver; the scsi-generic driver is using its own.
So we should move the current sense handling into the
scsi-disk driver.
Hannes Reinecke [Wed, 24 Nov 2010 11:15:57 +0000 (12:15 +0100)]
scsi: Return SAM status codes
Traditionally, the linux stack is using SCSI status codes
which are shifted by one as compared to those defined in SAM.
A SCSI emulation should naturally return the SAM defined codes,
not the linux ones.
So to avoid any confusion this patch modifies the existing
definitions to match those found in SAM and removes any
(now obsolete) byte-shift from the returned status codes.
Hannes Reinecke [Wed, 24 Nov 2010 11:15:56 +0000 (12:15 +0100)]
scsi: Increase the number of possible devices
The SCSI parallel interface has a limit of 8 devices, but
not the SCSI stack in general. So we should be removing the
hard-coded limit and use MAX_SCSI_DEVS instead.
And we only need to scan those devices which are allocated
by the bus.
Ryan Harper [Fri, 12 Nov 2010 17:07:13 +0000 (11:07 -0600)]
Implement drive_del to decouple block removal from device removal
Currently device hotplug removal code is tied to device removal via
ACPI. All pci devices that are removable via device_del() require the
guest to respond to the request. In some cases the guest may not
respond leaving the device still accessible to the guest. The management
layer doesn't currently have a reliable way to revoke access to host
resource in the presence of an uncooperative guest.
This patch implements a new monitor command, drive_del, which
provides an explicit command to revoke access to a host block device.
drive_del first quiesces the block device (qemu_aio_flush;
bdrv_flush() and bdrv_close()). This prevents further IO from being
submitted against the host device. Finally, drive_del cleans up
pointers between the drive object (host resource) and the device
object (guest resource).
Stefan Hajnoczi [Fri, 12 Nov 2010 09:57:11 +0000 (09:57 +0000)]
scsi-disk: Move active request asserts
SCSI read/write requests should not be re-issued before the current
fragment of I/O completes. There are asserts in scsi-disk.c that guard
this constraint but they trigger on SPARC Linux 2.4. It turns out that
the asserts are too early in the code path and don't allow for read
requests to terminate.
Only the read assert needs to be moved but move the write assert too for
consistency.