History | View | Annotate | Download (108.7 kB)
block: add .bdrv_co_write_zeroes() interface
The ability to zero regions of an image file is a useful primitive forhigher-level features such as image streaming or zero write detection.
Image formats may support an optimized metadata representation instead...
block: perform zero-detection during copy-on-read
Copy-on-Read populates the image file with data read from a backingimage. In order to avoid bloating the image file when all zeroes areread we should scan the buffer and perform an optimized zero write...
block: add bdrv_find_backing_image
Add bdrv_find_backing_image: given a BlockDriverState pointer, and an id,traverse the backing image chain to locate the id.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>...
block: check bdrv_in_use() before blockdev operations
Long-running block operations like block migration and image streamingmust have continual access to their block device. It is not safe toperform operations like hotplug, eject, change, resize, commit, or...
block: make copy-on-read a per-request flag
Previously copy-on-read could only be enabled for all requests to ablock device. This means requests coming from the guest as well asQEMU's internal requests would perform copy-on-read when enabled.
For image streaming we want to support finer-grained behavior than just...
block: add BlockJob interface for long-running operations
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>Signed-off-by: Kevin Wolf <kwolf@redhat.com>
block: avoid useless checks on acb->bh
Coverity is confused by this "if" and reports leaks on acb->bh.The bottom half is always deleted before releasing the AIOCB,in either bdrv_aio_cancel_em or bdrv_aio_bh_cb.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>...
block: bdrv_aio_* do not return NULL
Initially done with the following semantic patch:
rule1 expression E;statement S;@@ E =( bdrv_aio_readv | bdrv_aio_writev | bdrv_aio_flush | bdrv_aio_discard | bdrv_aio_ioctl) (...);(- if (E == NULL) { ... }...
rule1
block: simplify failure handling for bdrv_aio_multiwrite
Now that early failure of bdrv_aio_writev is not possible anymore,mcb->num_requests can be set before the loop starts.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>Signed-off-by: Kevin Wolf <kwolf@redhat.com>
block: convert qemu_aio_flush() calls to bdrv_drain_all()
Many places in QEMU call qemu_aio_flush() to complete all pendingasynchronous I/O. Most of these places actually want to drain all blockrequests but there is no block layer API to do so.
This patch introduces the bdrv_drain_all() API to wait for requests...
block: wait_for_overlapping_requests() deadlock detection
Debugging a reentrant request deadlock was fun but in the future we needa quick and obvious way of detecting such bugs. Add an assert thatchecks we are not about to deadlock when waiting for another request....
block: implement bdrv_co_is_allocated() boundary cases
Cases beyond the end of the disk image are only implemented for blockdrivers that do not provide .bdrv_co_is_allocated(). It's worth makingthese cases generic so that block drivers that do implement...
block: add request tracking
The block layer does not know about pending requests. This informationis necessary for copy-on-read since overlapping requests must beserialized to prevent races that corrupt the image.
The BlockDriverState gets a new tracked_request list field which...
block: add interface to toggle copy-on-read
The bdrv_enable_copy_on_read()/bdrv_disable_copy_on_read() functions canbe used to programmatically enable or disable copy-on-read for a blockdevice. Later patches add the actual copy-on-read logic.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>...
block: wait for overlapping requests
When copy-on-read is enabled it is necessary to wait for overlappingrequests before issuing new requests. This prevents races between thecopy-on-read and a write request.
block: request overlap detection
Detect overlapping requests and remember to align to cluster boundariesif the image format uses them. This assumes that allocating I/O isperformed in cluster granularity - which is true for qcow2, qed, etc.
block: core copy-on-read logic
block: drop .bdrv_is_allocated() interface
Now that all block drivers have been converted to.bdrv_co_is_allocated() we can drop .bdrv_is_allocated().
Note that the public bdrv_is_allocated() interface is still availablebut is in fact a synchronous wrapper around .bdrv_co_is_allocated()....
block: add bdrv_co_is_allocated() interface
This patch introduces the public bdrv_co_is_allocated() interface whichcan be used to query image allocation status while the VM is running.
block: use public bdrv_is_allocated() interface
There is no need for bdrv_commit() to use the BlockDriver.bdrv_is_allocated() interface directly. Converting to the publicinterface gives us the freedom to drop .bdrv_is_allocated() entirely infavor of a new .bdrv_co_is_allocated() in the future....
block: add .bdrv_co_is_allocated()
This patch adds the .bdrv_co_is_allocated() interface which is identicalto .bdrv_is_allocated() but runs in coroutine context. Running incoroutine context implies that other coroutines might be performing I/Oat the same time. Therefore it must be safe to run while the following...
block: add the blockio limits command line support
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>Signed-off-by: Kevin Wolf <kwolf@redhat.com>
block: add I/O throttling algorithm
hmp/qmp: add block_set_io_throttle
block: allow migration to work with image files (v3)
Image files have two types of data: immutable data that describes things likeimage size, backing files, etc. and mutable data that includes offset andreference count tables.
Today, image formats aggressively cache mutable data to improve performance. In...
block: Rename bdrv_co_flush to bdrv_co_flush_to_disk
There are two different types of flush that you can do: Flushing one level upto the OS (i.e. writing data to the host page cache) or flushing it all the waydown to the disk. The existing functions flush to the disk, reflect this in the...
block: Introduce bdrv_co_flush_to_os
qcow2 has a writeback metadata cache, so flushing a qcow2 image actuallyconsists of writing back that cache to the protocol and only then flushes theprotocol in order to get everything stable on disk.
This introduces a separate bdrv_co_flush_to_os to reflect the split....
block: Make cache=unsafe flush to the OS
cache=unsafe completely ignored bdrv_flush, because flushing the host diskcosts a lot of performance. However, this means that qcow2 images (andpotentially any other format) can lose data even after the guest has issued a...
block: add eject request callback
Recent versions of udev always keep the tray locked so that the kernelcan observe "eject request" events (aka tray button presses) even ondiscs that aren't mounted. Add support for these events in the ATAPIand SCSI cd drive device models....
Merge remote-tracking branch 'kwolf/for-anthony' into staging
Conflicts: block/vmdk.c
block: reinitialize across bdrv_close()/bdrv_open()
Several BlockDriverState fields are not being reinitialized acrossbdrv_close()/bdrv_open(). Make sure they are reset to their defaultvalues.
block: Remove dead code
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
block: Fix bdrv_open use after free
tmp_filename was used outside the block it was defined in, i.e. after it wentout of scope. Move its declaration to the top level.
block: set bs->read_only before .bdrv_open()
Several block drivers set bs->read_only in .bdrv_open() butblock.c:bdrv_open_common() clobbers its value. Additionally, QED usesbdrv_is_read_only() in .bdrv_open() to decide whether to performconsistency checks....
block: iostatus: Drop BDRV_IOS_INVAL
A future commit will convert bdrv_info() to the QAPI and it won'tprovide IOS_INVAL.
Luckily all we have to do is to add a new 'iostatus_enabled'member to BlockDriverState and use it instead.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
block: Rename the BlockIOStatus enum values
The biggest change is to rename its prefix from BDRV_IOS toBLOCK_DEVICE_IO_STATUS.
Next commit will convert the query-block command to the QAPIand that's how the enumeration is going to be generated.
qapi: Convert query-block
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
qapi: Convert query-blockstats
Drop qemu-objects.h from modules that don't require it
Previous commits dropped most qobjects usage from qemu modules(now they are a low level interface used by the QAPI). However,some modules still include the qemu-objects.h header file.
This commit drops qemu-objects.h from some of those modules...
block: change flush to co_flush
Since coroutine operation is now mandatory, convert all bdrv_flushimplementations to coroutines. For qcow2, this means taking the lock.Other implementations are simpler and just forward bdrv_flush to theunderlying protocol, so they can avoid the lock....
block: change discard to co_discard
Since coroutine operation is now mandatory, convert both bdrv_discardimplementations to coroutines. For qcow2, this means taking the lockaround the operation. raw-posix remains synchronous.
The bdrv_discard callback is then unused and can be eliminated....
block: unify flush implementations
Add coroutine support for flush and apply the same emulation thatwe already do for read/write. bdrv_aio_flush is simplified to alwaysgo through a coroutine.
block: add bdrv_co_discard and bdrv_aio_discard support
This similarly adds support for coroutine and asynchronous discard.
block: rename bdrv_co_rw_bh
block: drop emulation functions that use coroutines
Block drivers that implement coroutine functions used to get sync andaio wrappers. This is no longer necessary since all request processingnow happens in a coroutine. If a block driver implements the coroutine...
block: drop .bdrv_read()/.bdrv_write() emulation
There is no need to emulate .bdrv_read()/.bdrv_write() since theseinterfaces are only called if aio and coroutine interfaces are notpresent. All valid BlockDrivers must implement either sync, aio, orcoroutine interfaces....
block: drop bdrv_has_async_rw()
Commit cd74d83345e0e3b708330ab8c4cd9111bb82cda6 ("block: switchbdrv_read()/bdrv_write() to coroutines") removed the bdrv_has_async_rw()callers. This patch removes bdrv_has_async_rw() since it is no longerused.
block: switch bdrv_aio_readv() to coroutines
More sync, aio, and coroutine unification. Make bdrv_aio_readv() gothrough coroutine request processing.
block: mark blocks dirty on coroutine write completion
The aio write operation marks blocks dirty when the write operationcompletes. The coroutine write operation marks blocks dirty beforeissuing the write operation.
It seems safest to mark the block dirty when the operation completes so...
block: switch bdrv_aio_writev() to coroutines
More sync, aio, and coroutine unification. Make bdrv_aio_writev() gothrough coroutine request processing.
Remove the dirty block callback mechanism which was needed only for aioprocessing and can be done more naturally in coroutine context....
block: directly invoke .bdrv_* from emulation functions
The emulation functions which supply default BlockDriver .bdrv_*()functions given another implemented .bdrv_*() function should not usepublic bdrv_*() interfaces. This patch ensures they invoke .bdrv_*()...
block: split out bdrv_co_do_readv() and bdrv_co_do_writev()
The public interface for I/O in coroutine context is bdrv_co_readv() andbdrv_co_writev(). Split out the request processing code intobdrv_co_do_readv() and bdrv_co_writev() so that it can be called...
block: switch bdrv_read()/bdrv_write() to coroutines
The bdrv_read()/bdrv_write() functions call .bdrv_read()/.bdrv_write().They should go through bdrv_co_do_readv() and bdrv_co_do_writev()instead in order to unify request processing code across sync, aio, and...
block: directly invoke .bdrv_aio_*() in bdrv_co_io_em()
We will unify block layer request processing across sync, aio, andcoroutines and this means a .bdrv_co_*() emulation function should notcall back into the public interface. There's no need here, just call...
QMP: query-status: Add 'io-status' key
Contains the I/O status for the given device. The key is only presentif the device supports it and the VM is configured to stop on errors.
Please, check the documentation being added in this commit for moreinformation....
HMP: Print 'io-status' information
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>Reviewed-by: Markus Armbruster <armbru@redhat.com>Signed-off-by: Kevin Wolf <kwolf@redhat.com>
block: Keep track of devices' I/O status
This commit adds support to the BlockDriverState type to keep trackof devices' I/O status.
There are three possible status: BDRV_IOS_OK (no error), BDRV_IOS_ENOSPC(no space error) and BDRV_IOS_FAILED (any other error). The distinction...
trace: add arguments to bdrv_co_io_em() trace event
It is useful to know the BlockDriverState as well as thesector_num/nb_sectors of an emulated .bdrv_co_*() request.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
trace: trace bdrv_open_common()
bdrv_open_common() is a useful point to trace since it reveals thefilename and block driver for a given BlockDriverState.
block: New bdrv_set_buffer_alignment()
Device models should be able to set it without an unclean include ofblock_int.h.
Signed-off-by: Markus Armbruster <armbru@redhat.com>Signed-off-by: Kevin Wolf <kwolf@redhat.com>
block: Reset buffer alignment on detach
BlockDriverState member buffer_alignment is initially 512. The devicemodel may set them, with bdrv_set_buffer_alignment(). If the devicemodel gets detached (hot unplug), the device's alignment is leftbehind. Only okay because device hot unplug automatically destroys...
block: New change_media_cb() parameter load
To let device models distinguish between eject and load.
block: Clean up remaining users of "removable"
BlockDriverState member removable is a confused mess. It is true whenan ide-cd, scsi-cd or floppy qdev is attached, or when theBlockDriverState was created with -drive if={floppy,sd} or -driveif={ide,scsi,xen,none},media=cdrom ("created removable"), except when...
block: Drop BlockDriverState member removable
It's a confused mess (see previous commit). No users remain.
block: Show whether the virtual tray is open in info block
Need to ask the device, so this requires new BlockDevOps memberis_tray_open().
block: Revert entanglement of bdrv_is_inserted() with tray status
Commit 4be9762a changed bdrv_is_inserted() to fail when the tray isopen. Unfortunately, there are two different kinds of users, withconflicting needs.
1. Device models using bdrv_eject(), currently ide-cd and scsi-cd....
block: Drop tray status tracking, no longer used
Commit 4be9762a is now completely redone.
block: Leave enforcing tray lock to device models
The device model knows best when to accept the guest's eject command.No need to detour through the block layer.
bdrv_eject() can't fail anymore. Make it void.
Signed-off-by: Markus Armbruster <armbru@redhat.com>...
block: Drop medium lock tracking, ask device models instead
Requires new BlockDevOps member is_medium_locked(). Implement for IDEand SCSI CD-ROMs.
block: Rename bdrv_set_locked() to bdrv_lock_medium()
While there, make the locked parameter bool.
block: Clean up bdrv_flush_all()
Change (!bdrv_is_removable(bs) || bdrv_is_inserted(bs)) to justbdrv_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...
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 previouscheckpoint will not roll back writes to block devices with removable...
block: Leave tracking media change to device models
hw/fdc.c is the only one that cares.
block: Attach non-qdev devices as well
For now, this just protects against programming errors like having thesame 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()...
block: Generalize change_cb() to BlockDevOps
So we can more easily add device model callbacks.
block: Split change_cb() into change_media_cb(), resize_cb()
Multiplexing callbacks complicates matters needlessly.
Use new macro QEMU_PACKED for packed structures
Most changes were made using these commands:
git grep -la '__attribute__((packed))'|xargs perl -pi -e 's/__attribute__\(\(packed\)\)/QEMU_PACKED/'git grep -la '__attribute__ ((packed))'|xargs perl -pi -e 's/__attribute__ \(\(packed\)\)/QEMU_PACKED/'...
block: latency accounting
Account the total latency for read/write/flush requests. This allowsmanagement tools to average it based on a snapshot of the nr opscounters and allow checking for SLAs or provide statistics.
Signed-off-by: Christoph Hellwig <hch@lst.de>...
block: explicit I/O accounting
Decouple the I/O accounting from bdrv_aio_readv/writev/flush andmake the hardware models call directly into the accounting helpers.
This means: - we do not count internal requests from image formats in addition to guest originating I/O...
block: include flush requests in info blockstats
Signed-off-by: Christoph Hellwig <hch@lst.de>Signed-off-by: Kevin Wolf <kwolf@redhat.com>
block: parse cache mode flags in a single place
This patch introduces bdrv_parse_cache_flags() which sets open flagsgiven a cache mode. Previously this was duplicated in blockdev.c andqemu-img.c.
block: add cache=directsync parameter to -drive
This patch adds -drive cache=directsync for O_DIRECT | O_SYNC host fileI/O with no disk write cache presented to the guest.
This mode is useful when guests may not be sending flushes whenappropriate and therefore leave data at risk in case of power failure....
fix code format
Fix code format to make checkpatch.pl happy.
Signed-off-by: Robert Wang <wdongxu@linux.vnet.ibm.com>Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Use glib memory allocation and free functions
qemu_malloc/qemu_free no longer exist after this commit.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
block: Use bdrv_co_* instead of synchronous versions in coroutines
If we're already in a coroutine, there is no reason to use the synchronousversion of block layer functions when a coroutine one exists. This makesbdrv_read/write/flush use bdrv_co_* when used inside a coroutine....
async: Remove AsyncContext
The purpose of AsyncContexts was to protect qcow and qcow2 against reentrancyduring an emulated bdrv_read/write (which includes a qemu_aio_wait() call andcan run AIO callbacks of different requests if it weren't for AsyncContexts)....
block: Add bdrv_co_readv/writev
Add new block driver callbacks bdrv_co_readv/writev, which work on aQEMUIOVector like bdrv_aio_*, but don't need a callback. The function may onlybe called inside a coroutine, so a block driver implementing this interface can...
block: Emulate AIO functions with bdrv_co_readv/writev
Use the bdrv_co_readv/writev callbacks to implement bdrv_aio_readv/writev andbdrv_read/write if a driver provides the coroutine version instead of thesynchronous or AIO version.
block: Add bdrv_co_readv/writev emulation
In order to be able to call bdrv_co_readv/writev for drivers that don'timplement the functions natively, add an emulation that uses the AIO functionsto implement them.
block: Removed unused function bdrv_write_sync
Signed-off-by: Frediano Ziglio <freddy77@gmail.com>Signed-off-by: Kevin Wolf <kwolf@redhat.com>
block: Make BlockDriver method bdrv_eject() return void
Callees always return 0, except for FreeBSD's cdrom_eject(), whichreturns -ENOTSUP when the device is in a terminally wedged state.
The only caller is bdrv_eject(), and it maps -ENOTSUP to 0 since...
block: Don't let locked flag prevent medium load
Commit aea2a33c made bdrv_eject() obey the locked flag. Correct formedium eject (eject_flag set), incorrect for medium load (eject_flagclear). See MMC-5 Table 341 "Actions for Lock/Unlock/Eject".
block: Reset device model callbacks on detach
BlockDriverState members change_cb and change_opaque are initiallynull. The device model may set them, with bdrv_set_change_cb(). Ifthe device model gets detached (hot unplug), they're left dangling.Only safe because device hot unplug automatically destroys the...
block: add bdrv_get_allocated_file_size() operation
qemu-img.c wants to count allocated file size of image. Previously itcounts a single bs->file by 'stat' or Window API. As VMDK introducesmultiple file support, the operation becomes format specific with...
bdrv_img_create: Fix segfault
Block drivers that don't support creating images don't have a size option. Failgracefully instead of segfaulting when trying to access the option's value.
block: clarify the meaning of BDRV_O_NOCACHE
Change BDRV_O_NOCACHE to only imply bypassing the host OS file cache,but no writeback semantics. All existing callers are changed to alsospecify BDRV_O_CACHE_WB to give them writeback semantics.
block: Remove type hint, it's guest matter, doesn't belong here
No users of bdrv_get_type_hint() left. bdrv_set_type_hint() can makethe media removable by side effect. Make that explicit.
block QMP: Deprecate query-block's "type", drop info block's "type="
query-block's specification documents response member "type" withvalues "hd", "cdrom", "floppy", "unknown".
Its value is unreliable: a block device used as floppy has type"floppy" if created with if=floppy, but type "hd" if created with...
Fix typos in comments and code (occured -> occurred and related)
The code changed here is an unused data type name (evt_flush_occurred).
Signed-off-by: Stefan Weil <weil@mail.berlios.de>Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Fix typo in code and comments
Replace writeable -> writable