Not that you can request a >2GiB transaction, but that's why checking
for it makes no sense anymore.
With the newer 'limit' parameter to prepare_buf, we no longer need a
static limit. The maximum limit is still 2GiB, but the limit parameter
is set to the current transaction size, which cannot surpass 32MiB
(512 * 65536). If the PRDT surpasses the transactional size, then,
we'll just carry out the normative underflow handling pathways instead
of needing an extra, strange pathway that worries about hitting some
logistical cap for the largest sglist we can support -- we'll never
even attempt to build one that big anymore.
Reported-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1445902682-20051-1-git-send-email-jsnow@redhat.com
prepare_buf should not always grab as many descriptors
as it can, sometimes it should self-limit.
For example, an NCQ transfer of 1 sector with a PRDT that
describes 4GiB of data should not copy 4GiB of data, it
should just transfer that first 512 bytes.
PIO is not affected, because the dma_buf_rw dma helpers
already have a byte limit built-in to them, but DMA/NCQ
will exhaust the entire list regardless of requested size.
AHCI 1.3 specifies in section 6.1.6 Command List Underflow that
NCQ is not required to detect underflow conditions. Non-NCQ
pathways signal underflow by writing to the PRDBC field, which
will already occur by writing the actual transferred byte count
to the PRDBC, signaling the underflow.
Our NCQ pathways aren't required to detect underflow, but since our DMA
backend uses the size of the PRDT to determine the size of the transer,
if our PRDT is bigger than the transaction (the underflow condition) it
doesn't cost us anything to detect it and truncate the PRDT.
This is a recoverable error and is not signaled to the guest, in either
NCQ or normal DMA cases.
For BMDMA, the existing pathways should see no guest-visible difference,
but any bytes described in the overage will no longer be transferred
before indicating to the guest that there was an underflow.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1435767578-32743-2-git-send-email-jsnow@redhat.com
We create optional sections with this patch. But we already have
optional subsections. Instead of having two mechanism that do the
same, we can just generalize it.
For subsections we just change:
- Add a needed function to VMStateDescription
- Remove VMStateSubsection (after removal of the needed function
it is just a VMStateDescription)
- Adjust the whole tree, moving the needed function to the corresponding
VMStateDescription
Signed-off-by: Juan Quintela <quintela@redhat.com>
valgrind complains about:
==16447== 16 bytes in 2 blocks are definitely lost in loss record 1,304 of 3,310
==16447== at 0x4C2845D: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==16447== by 0x2E4FD7: malloc_and_trace (vl.c:2546)
==16447== by 0x64C770E: g_malloc (in /usr/lib64/libglib-2.0.so.0.3600.3)
==16447== by 0x36FB47: qemu_extend_irqs (irq.c:55)
==16447== by 0x36FBD3: qemu_allocate_irqs (irq.c:64)
==16447== by 0x3B4B44: bmdma_init (pci.c:464)
==16447== by 0x3B547B: pci_piix_init_ports (piix.c:144)
==16447== by 0x3B55D2: pci_piix_ide_realize (piix.c:164)
==16447== by 0x3EAEC6: pci_qdev_realize (pci.c:1790)
==16447== by 0x36C685: device_set_realized (qdev.c:1058)
==16447== by 0x47179E: property_set_bool (object.c:1514)
==16447== by 0x470098: object_property_set (object.c:837)
Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com>
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)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=hlbC
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Block patches for 2.3
# gpg: Signature made Tue Mar 10 13:03:17 2015 GMT using RSA key ID C88F2FD6
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>"
* remotes/kevin/tags/for-upstream: (73 commits)
MAINTAINERS: Add jcody as blockjobs, block devices maintainer
iotests: add O_DIRECT alignment probing test
block/raw-posix: fix launching with failed disks
MAINTAINERS: Add jsnow as IDE maintainer
sheepdog: Fix misleading error messages in sd_snapshot_create()
Add testcase for scsi-hd devices without drive property
scsi-hd: fix property unset case
block/vdi: Add locking for parallel requests
iotests: Drop vpc from 004's and 104's format list
iotests: Remove 006
iotests: Fix 051's reference output
virtio-blk: Remove the stale FIXME comment
tests: Check QVIRTIO_F_ANY_LAYOUT flag in virtio-blk test
libqos: Solve bug in interrupt checking when using MSIX in virtio-pci.c
sheepdog: fix confused return values
qtest/ahci: add fragmented dma test
qtest/ahci: Add PIO and LBA48 tests
qtest/ahci: Add DMA test variants
libqos/ahci: add ahci command helpers
qtest/ahci: Add a macro bootup routine
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This moves more common restarting logic to the core IDE code.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1424708286-16483-10-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Start moving the initial state of the current request to IDEBus, so that
AHCI can use it. The set_unit callback is not used anymore once this is
done.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1424708286-16483-9-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
With BMDMA specific excised from the restart functions,
create a HBA-agnostic restart callback to be shared
between the different HBAs.
Change the callback registered with the vmstate_change
handler to always point to ide_restart_cb instead of
relying on the IDEDMAOps.restart_cb() member.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1424708286-16483-7-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Pass the containing IDEBus to the restart_cb instead
of the more specific BMDMAState child.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1424708286-16483-6-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Whenever an error stops the VM, ide_handle_rw_error does
"s->bus->dma->unit = s->unit". So we can just use
idebus_active_if.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1424708286-16483-5-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch adds the restart_dma callback and adjusts
the ide_restart_dma function to utilize this callback
to call the BMDMA-specific restart code instead of statically
executing BMDMA-specific code.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1424708286-16483-3-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch begins refactoring the restart dma functions
out of bmdma to be shared with AHCI and other future
IDE HBA implementations.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1424708286-16483-2-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(With the previous atapi_dma flag recovery)
If migration happens between the ATAPI command being written and the
bmdma being started, the DMA is dropped. Eventually the guest times
out and recovers, but that can take many seconds.
(This is rare, on a pingpong reading the CD continuously I hit
this about ~1/30-1/50 migrates)
I don't think we've got enough state to be able to recover safely
at this point, so I throw a 'medium error, no seek complete'
that I'm assuming guests will try and recover from an apparently
dirty CD.
OK, it's a hack, the real solution is probably to push a lot of
ATAPI state into the migration stream, but this is a fix that
works with no stream changes. Tested only on Linux (both RHEL5
(pre-libata) and RHEL7).
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This impacts both BMDMA and AHCI HBA interfaces for IDE.
Currently, we confuse the difference between a PRDT having
"0 bytes" and a PRDT having "0 complete sectors."
When we receive an incomplete sector, inconsistent error checking
leads to an infinite loop wherein the call succeeds, but it
didn't give us enough bytes -- leading us to re-call the
DMA chain over and over again. This leads to, in the BMDMA case,
leaked memory for short PRDTs, and infinite loops and resource
usage in the AHCI case.
The .prepare_buf() callback is reworked to return the number of
bytes that it successfully prepared. 0 is a valid, non-error
answer that means the table was empty and described no bytes.
-1 indicates an error.
Our current implementation uses the io_buffer in IDEState to
ultimately describe the size of a prepared scatter-gather list.
Even though the AHCI PRDT/SGList can be as large as 256GiB, the
AHCI command header limits transactions to just 4GiB. ATA8-ACS3,
however, defines the largest transaction to be an LBA48 command
that transfers 65,536 sectors. With a 512 byte sector size, this
is just 32MiB.
Since our current state structures use the int type to describe
the size of the buffer, and this state is migrated as int32, we
are limited to describing 2GiB buffer sizes unless we change the
migration protocol.
For this reason, this patch begins to unify the assertions in the
IDE pathways that the scatter-gather list provided by either the
AHCI PRDT or the PCI BMDMA PRDs can only describe, at a maximum,
2GiB. This should be resilient enough unless we need a sector
size that exceeds 32KiB.
Further, the likelihood of any guest operating system actually
attempting to transfer this much data in a single operation is
very slim.
To this end, the IDEState variables have been updated to more
explicitly clarify our maximum supported size. Callers to the
prepare_buf callback have been reworked to understand the new
return code, and all versions of the prepare_buf callback have
been adjusted accordingly.
Lastly, the ahci_populate_sglist helper, relied upon by the
AHCI implementation of .prepare_buf() as well as the PCI
implementation of the callback have had overflow assertions
added to help make clear the reasonings behind the various
type changes.
[Added %d -> %"PRId64" fix John sent because off_pos changed from int to
int64_t.
--Stefan]
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1414785819-26209-4-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Device models should access their block backends only through the
block-backend.h API. Convert them, and drop direct includes of
inappropriate headers.
Just four uses of BlockDriverState are left:
* The Xen paravirtual block device backend (xen_disk.c) opens images
itself when set up via xenbus, bypassing blockdev.c. I figure it
should go through qmp_blockdev_add() instead.
* Device model "usb-storage" prompts for keys. No other device model
does, and this one probably shouldn't do it, either.
* ide_issue_trim_cb() uses bdrv_aio_discard() instead of
blk_aio_discard() because it fishes its backend out of a BlockAIOCB,
which has only the BlockDriverState.
* PC87312State has an unused BlockDriverState[] member.
The next two commits take care of the latter two.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
I'll use it with block backends shortly, and the name is going to fit
badly there. It's a block layer thing anyway, not just a block driver
thing.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
They are not used by AHCI, and should not be even available there.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
It is now called only after the set_inactive callback. Put the two together.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Drop the unused return value and make the callback optional.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Drop the unused return value and make the callback optional.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Drop the unused return value and make the callback optional.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
After previous Peter patch, they are redundant. This way we don't
assign them except when needed. Once there, there were lots of case
where the ".fields" indentation was wrong:
.fields = (VMStateField []) {
and
.fields = (VMStateField []) {
Change all the combinations to:
.fields = (VMStateField[]){
The biggest problem (appart from aesthetics) was that checkpatch complained
when we copy&pasted the code from one place to another.
Signed-off-by: Juan Quintela <quintela@redhat.com>
Acked-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
This fixes a warning from the static code analysis (smatch).
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
The device is supposed to reset the Bus Master IDE Active bit in the
status register when 0 is written to the Start/Stop Bus Master bit in
the command register.
In the common cases this happens automatically because bdrv_drain_all()
flushes the requests, but with a large PRDT it could remain set.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Many of these should be cleaned up with proper qdev-/QOM-ification.
Right now there are many catch-all headers in include/hw/ARCH depending
on cpu.h, and this makes it necessary to compile these files per-target.
However, fixing this does not belong in these patches.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The Bus Master IDE Active bit (BM_STATUS_DMAING) is not only set when
the request is still in flight, but also when it has completed and the
size of the physical memory regions in the PRDT was larger than the
transfer size.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
target_phys_addr_t is unwieldly, violates the C standard (_t suffixes are
reserved) and its purpose doesn't match the name (most target_phys_addr_t
addresses are not target specific). Replace it with a finger-friendly,
standards conformant hwaddr.
Outstanding patchsets can be fixed up with the command
git rebase -i --exec 'find -name "*.[ch]"
| xargs s/target_phys_addr_t/hwaddr/g' origin
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Do this while we are touching this part of the code, before introducing
more uses of "int is_read".
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Back when I made patches introducing dma_addr_t and various PCI DMA
wrapper functions, I made a mistake. The bmdma_addr_{read,write} functions
need to take target_phys_addr_t not dma_addr_t, since they are assigned
to MemoryRegionOps callbacks.
This patch corrects my error.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Alexander Graf <agraf@suse.de>
This patch removes some unnecessary casts in the PCI IDE device,
introduced by commit 552908fef5
'PCI IDE: Use PCI DMA stub functions'.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Many places in QEMU call qemu_aio_flush() to complete all pending
asynchronous I/O. Most of these places actually want to drain all block
requests but there is no block layer API to do so.
This patch introduces the bdrv_drain_all() API to wait for requests
across all BlockDriverStates to complete. As a bonus we perform checks
after qemu_aio_wait() to ensure that requests really have finished.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This updates the PCI IDE device emulation to use the explicit PCI DMA
wrapper to initialize its scatter/gathjer structure. This means this
driver should not need further changes when the sglist interface is
extended to support IOMMUs.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Today, when notifying a VM state change with vm_state_notify(),
we pass a VMSTOP macro as the 'reason' argument. This is not ideal
because the VMSTOP macros tell why qemu stopped and not exactly
what the current VM state is.
One example to demonstrate this problem is that vm_start() calls
vm_state_notify() with reason=0, which turns out to be VMSTOP_USER.
This commit fixes that by replacing the VMSTOP macros with a proper
state type called RunState.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Including it in device models is unclean, including it without a
reason adds insult to injury.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Clearing the error status flag was missing for restarting flushes. Now that the
error status is separate from the BM status register, we can simply set it to 0
after restarting the request. This ensures that we never forget to clear a bit.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add support for TRIM sub function of the data set management command,
and wire it up to the qemu discard infrastructure.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Replace the is_read flag with a dma_cmd flag to allow the dma and
restart logic to handler other commands like TRIM.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When adding the werror=stop mode, some flags were added to s->status
which are used to determine what kind of operation should be restarted
when the VM is continued.
Unfortunately, it turns out that s->status is in fact a device register
and as such is visible to the guest (some of the abused bits are even
writable for the guest).
For migration we keep on using the old VMState field (renamed to
migration_compat_status) if the status register doesn't use any of the
previously abused bits. If it does, we use a subsection with a clean copy of
the status register.
The error status is always sent in a subsection if there is any error. It can't
use the old field because errors happen even without PCI.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
These printfs aren't really debug messages, but clearly indicate a bug if they
ever become effective. Noone uses DEBUG_IDE, let's re-enable the check
unconditionally and make it an assertion instead of printfs in the device
emulation.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
With bm == NULL, other code in the same function would crash.
This bug was reported by cppcheck:
hw/ide/pci.c:280: error: Possible null pointer dereference: bm
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
John Snow