Leonid Bloch [Wed, 26 Sep 2018 16:04:43 +0000 (19:04 +0300)]
qcow2: Assign the L2 cache relatively to the image size
Sufficient L2 cache can noticeably improve the performance when using
large images with frequent I/O.
Previously, unless 'cache-size' was specified and was large enough, the
L2 cache was set to a certain size without taking the virtual image size
into account.
Now, the L2 cache assignment is aware of the virtual size of the image,
and will cover the entire image, unless the cache size needed for that is
larger than a certain maximum. This maximum is set to 1 MB by default
(enough to cover an 8 GB image with the default cluster size) but can
be increased or decreased using the 'l2-cache-size' option. This option
was previously documented as the *maximum* L2 cache size, and this patch
makes it behave as such, instead of as a constant size. Also, the
existing option 'cache-size' can limit the sum of both L2 and refcount
caches, as previously.
Leonid Bloch [Wed, 26 Sep 2018 16:04:42 +0000 (19:04 +0300)]
qcow2: Avoid duplication in setting the refcount cache size
The refcount cache size does not need to be set to its minimum value in
read_cache_sizes(), as it is set to at least its minimum value in
qcow2_update_options_prepare().
Leonid Bloch [Wed, 26 Sep 2018 16:04:40 +0000 (19:04 +0300)]
include: Add a lookup table of sizes
Adding a lookup table for the powers of two, with the appropriate size
prefixes. This is needed when a size has to be stringified, in which
case something like '(1 * KiB)' would become a literal '(1 * (1L << 10))'
string. Powers of two are used very often for sizes, so such a table
will also make it easier and more intuitive to write them.
This table is generatred using the following AWK script:
Alberto Garcia [Thu, 6 Sep 2018 09:37:09 +0000 (12:37 +0300)]
block: Allow changing 'detect-zeroes' on reopen
'detect-zeroes' is one of the basic BlockdevOptions available for all
drivers, but it's not handled by bdrv_reopen_prepare(), so any attempt
to change it results in an error:
Alberto Garcia [Thu, 6 Sep 2018 09:37:08 +0000 (12:37 +0300)]
block: Allow changing 'discard' on reopen
'discard' is one of the basic BlockdevOptions available for all
drivers, but it's not handled by bdrv_reopen_prepare() so any attempt
to change it results in an error:
Alberto Garcia [Thu, 6 Sep 2018 09:37:07 +0000 (12:37 +0300)]
file-posix: Forbid trying to change unsupported options during reopen
The file-posix code is used for the "file", "host_device" and
"host_cdrom" drivers, and it allows reopening images. However the only
option that is actually processed is "x-check-cache-dropped", and
changes in all other options (e.g. "filename") are silently ignored:
While we could allow changing some of the other options, let's keep
things as they are for now but return an error if the user tries to
change any of them.
Alberto Garcia [Thu, 6 Sep 2018 09:37:06 +0000 (12:37 +0300)]
block: Forbid trying to change unsupported options during reopen
The bdrv_reopen_prepare() function checks all options passed to each
BlockDriverState (in the reopen_state->options QDict) and makes all
necessary preparations to apply the option changes requested by the
user.
Options are removed from the QDict as they are processed, so at the
end of bdrv_reopen_prepare() only the options that can't be changed
are left. Then a loop goes over all remaining options and verifies
that the old and new values are identical, returning an error if
they're not.
The problem is that at the moment there are options that are removed
from the QDict although they can't be changed. The consequence of this
is any modification to any of those options is silently ignored:
(qemu) qemu-io virtio0 "reopen -o discard=on"
This happens when all options from bdrv_runtime_opts are removed
from the QDict but then only a few of them are processed. Since
it's especially important that "node-name" and "driver" are not
changed, the code puts them back into the QDict so they are checked
at the end of the function. Instead of putting only those two options
back into the QDict, this patch puts all unprocessed options using
qemu_opts_to_qdict().
update_flags_from_options() also needs to be modified to prevent
BDRV_OPT_CACHE_NO_FLUSH, BDRV_OPT_CACHE_DIRECT and BDRV_OPT_READ_ONLY
from going back to the QDict.
Alberto Garcia [Thu, 6 Sep 2018 09:37:05 +0000 (12:37 +0300)]
block: Allow child references on reopen
In the previous patches we removed all child references from
bs->{options,explicit_options} because keeping them is useless and
wrong.
Because of this, any attempt to reopen a BlockDriverState using a
child reference as one of its options would result in a failure,
because bdrv_reopen_prepare() would detect that there's a new option
(the child reference) that wasn't present in bs->options.
But passing child references on reopen can be useful. It's a way to
specify a BDS's child without having to pass recursively all of the
child's options, and if the reference points to a different BDS then
this can allow us to replace the child.
However, replacing the child is something that needs to be implemented
case by case and only when it makes sense. For now, this patch allows
passing a child reference as long as it points to the current child of
the BlockDriverState.
It's also important to remember that, as a consequence of the
previous patches, this child reference will be removed from
bs->{options,explicit_options} after the reopening has been completed.
Alberto Garcia [Thu, 6 Sep 2018 09:37:03 +0000 (12:37 +0300)]
block: Remove child references from bs->{options,explicit_options}
Block drivers allow opening their children using a reference to an
existing BlockDriverState. These references remain stored in the
'options' and 'explicit_options' QDicts, but we don't need to keep
them once everything is open.
What is more important, these values can become wrong if the children
change:
Alberto Garcia [Thu, 6 Sep 2018 09:37:02 +0000 (12:37 +0300)]
file-posix: x-check-cache-dropped should default to false on reopen
The default value of x-check-cache-dropped is false. There's no reason
to use the previous value as a default in raw_reopen_prepare() because
bdrv_reopen_queue_child() already takes care of putting the old
options in the BDRVReopenState.options QDict.
If x-check-cache-dropped was previously set but is now missing from
the reopen QDict then it should be reset to false.
Alberto Garcia [Thu, 6 Sep 2018 09:37:01 +0000 (12:37 +0300)]
qemu-io: Fix writethrough check in reopen
"qemu-io reopen" doesn't allow changing the writethrough setting of
the cache, but the check is wrong, causing an error even on a simple
reopen with the default parameters:
Peter Maydell [Fri, 28 Sep 2018 17:56:09 +0000 (18:56 +0100)]
Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20180926' into staging
Queued tcg patches
# gpg: Signature made Wed 26 Sep 2018 19:27:22 BST
# gpg: using RSA key 64DF38E8AF7E215F
# gpg: Good signature from "Richard Henderson <[email protected]>"
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A 05C0 64DF 38E8 AF7E 215F
* remotes/rth/tags/pull-tcg-20180926:
tcg/i386: fix vector operations on 32-bit hosts
qht-bench: add -p flag to precompute hash values
qht: constify arguments to some internal functions
qht: constify qht_statistics_init
qht: constify qht_lookup
qht: fix comment in qht_bucket_remove_entry
qht: drop ht argument from qht iterators
test-qht: speed up + test qht_resize
test-qht: test deletion of the last entry in a bucket
test-qht: test removal of non-existent entries
test-qht: test qht_iter_remove
qht: add qht_iter_remove
qht: remove unused map param from qht_remove__locked
Peter Maydell [Fri, 28 Sep 2018 16:07:23 +0000 (17:07 +0100)]
Merge remote-tracking branch 'remotes/dgilbert/tags/pull-migration-20180926a' into staging
Migration pull 2018-09-26
This supercedes Juan's pull from the 13th
# gpg: Signature made Wed 26 Sep 2018 18:07:30 BST
# gpg: using RSA key 0516331EBC5BFDE7
# gpg: Good signature from "Dr. David Alan Gilbert (RH2) <[email protected]>"
# Primary key fingerprint: 45F5 C71B 4A0C B7FB 977A 9FA9 0516 331E BC5B FDE7
* remotes/dgilbert/tags/pull-migration-20180926a:
migration/ram.c: Avoid taking address of fields in packed MultiFDInit_t struct
migration: fix the compression code
migration: fix QEMUFile leak
tests/migration: Speed up the test on ppc64
migration: cleanup in error paths in loadvm
migration/postcopy: Clear have_listen_thread
tests/migration: Add migration-test header file
tests/migration: Support cross compilation in generating boot header file
tests/migration: Convert x86 boot block compilation script into Makefile
migration: use save_page_use_compression in flush_compressed_data
migration: show the statistics of compression
migration: do not flush_compressed_data at the end of iteration
Add a hint message to loadvm and exits on failure
migration: handle the error condition properly
migration: fix calculating xbzrle_counters.cache_miss_rate
migration/rdma: Fix uninitialised rdma_return_path
Peter Maydell [Fri, 28 Sep 2018 12:35:26 +0000 (13:35 +0100)]
Merge remote-tracking branch 'remotes/famz/tags/staging-pull-request' into staging
Block and testing patches
- Paolo's AIO fixes.
- VMDK streamOptimized corner case fix
- VM testing improvment on -cpu
# gpg: Signature made Wed 26 Sep 2018 03:54:08 BST
# gpg: using RSA key CA35624C6A9171C6
# gpg: Good signature from "Fam Zheng <[email protected]>"
# Primary key fingerprint: 5003 7CB7 9706 0F76 F021 AD56 CA35 624C 6A91 71C6
* remotes/famz/tags/staging-pull-request:
vmdk: align end of file to a sector boundary
tests/vm: Use -cpu max rather than -cpu host
aio-posix: do skip system call if ctx->notifier polling succeeds
aio-posix: compute timeout before polling
aio-posix: fix concurrent access to poll_disable_cnt
Peter Maydell [Fri, 28 Sep 2018 10:22:36 +0000 (11:22 +0100)]
Merge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-3.1-pull-request' into staging
- some fixes for setrlimit() and write()
- fixes ELF loader when host page size is greater than target page size
- add SO_LINGER to getsockopt()/setsockopt()
- move TargetFdTrans from syscall.c
v2: add "#include <linux/netlink.h>" in linux-user/fd-trans.c
* remotes/vivier2/tags/linux-user-for-3.1-pull-request:
linux-user: do setrlimit selectively
linux-user: write(fd, NULL, 0) parity with linux's treatment of same
linux-user: elf: mmap all the target-pages of hostpage for data segment
linux-user: add SO_LINGER to {g,s}etsockopt
linux-user: move TargetFdTrans functions to their own file
Peter Maydell [Tue, 25 Sep 2018 16:19:24 +0000 (17:19 +0100)]
migration/ram.c: Avoid taking address of fields in packed MultiFDInit_t struct
Taking the address of a field in a packed struct is a bad idea, because
it might not be actually aligned enough for that pointer type (and
thus cause a crash on dereference on some host architectures). Newer
versions of clang warn about this:
migration/ram.c:651:19: warning: taking address of packed member 'magic' of class or structure 'MultiFDInit_t' may result in an unaligned pointer value [-Waddress-of-packed-member]
migration/ram.c:652:19: warning: taking address of packed member 'version' of class or structure 'MultiFDInit_t' may result in an unaligned pointer value [-Waddress-of-packed-member]
migration/ram.c:737:19: warning: taking address of packed member 'magic' of class or structure 'MultiFDPacket_t' may result in an unaligned pointer value [-Waddress-of-packed-member]
migration/ram.c:745:19: warning: taking address of packed member 'version' of class or structure 'MultiFDPacket_t' may result in an unaligned pointer value [-Waddress-of-packed-member]
migration/ram.c:755:19: warning: taking address of packed member 'size' of class or structure 'MultiFDPacket_t' may result in an unaligned pointer value [-Waddress-of-packed-member]
Avoid the bug by not using the "modify in place" byteswapping
functions.
Fei Li [Tue, 25 Sep 2018 09:14:40 +0000 (17:14 +0800)]
migration: fix the compression code
Add judgement in compress_threads_save_cleanup() to check whether the
static CompressParam *comp_param has been allocated. If not, just
return; or else segmentation fault will occur when using the NULL
comp_param's parameters. One test case can reproduce this is: set
the compression on and migrate to a wrong nonexistent host IP address.
Our current code does not judge before handling comp_param[idx]'s quit
and cond that whether they have been initialized. If not initialized,
"qemu_mutex_lock_impl: Assertion `mutex->initialized' failed." will
occur. Fix this by squashing the terminate_compression_threads() into
compress_threads_save_cleanup() and employing the existing judgement
condition. One test case can reproduce this error is: set the
compression on and fail to fully setup the default eight compression
thread in compress_threads_save_setup().
Direct leak of 33864 byte(s) in 1 object(s) allocated from:
#0 0x7f3da7f31e50 in calloc (/lib64/libasan.so.5+0xeee50)
#1 0x7f3da644441d in g_malloc0 (/lib64/libglib-2.0.so.0+0x5241d)
#2 0x55af9db15440 in qemu_fopen_channel_input /home/elmarco/src/qemu/migration/qemu-file-channel.c:183
#3 0x55af9db15413 in channel_get_output_return_path /home/elmarco/src/qemu/migration/qemu-file-channel.c:159
#4 0x55af9db0d4ac in qemu_file_get_return_path /home/elmarco/src/qemu/migration/qemu-file.c:78
#5 0x55af9dad5e4f in open_return_path_on_source /home/elmarco/src/qemu/migration/migration.c:2295
#6 0x55af9dadb3bf in migrate_fd_connect /home/elmarco/src/qemu/migration/migration.c:3111
#7 0x55af9dae1bf3 in migration_channel_connect /home/elmarco/src/qemu/migration/channel.c:91
#8 0x55af9daddeca in socket_outgoing_migration /home/elmarco/src/qemu/migration/socket.c:108
#9 0x55af9e13d3db in qio_task_complete /home/elmarco/src/qemu/io/task.c:158
#10 0x55af9e13ca03 in qio_task_thread_result /home/elmarco/src/qemu/io/task.c:89
#11 0x7f3da643b1ca in g_idle_dispatch gmain.c:5535
Thomas Huth [Mon, 17 Sep 2018 17:12:10 +0000 (19:12 +0200)]
tests/migration: Speed up the test on ppc64
The SLOF boot process is always quite slow ... but we can speed it up
a little bit by specifying "-nodefaults" and by using the "nvramrc"
variable instead of "boot-command" (since "nvramrc" is evaluated earlier
in the SLOF boot process than "boot-command").
There's a couple of error paths in qemu_loadvm_state
which happen early on but after we've initialised the
load state; that needs to be cleaned up otherwise
we can hit asserts if the state gets reinitialised later.
Clear have_listen_thread when we exit the thread.
The fallout from this was that various things thought there was
an ongoing postcopy after the postcopy had finished.
The case that failed was postcopy->savevm->loadvm.
This corresponds to RH bug https://bugzilla.redhat.com/show_bug.cgi?id=1608765
Roman Kapl [Fri, 24 Aug 2018 13:17:34 +0000 (15:17 +0200)]
tcg/i386: fix vector operations on 32-bit hosts
The TCG backend uses LOWREGMASK to get the low 3 bits of register numbers.
This was defined as no-op for 32-bit x86, with the assumption that we have
eight registers anyway. This assumption is not true once we have xmm regs.
Since LOWREGMASK was a no-op, xmm register indidices were wrong in opcodes
and have overflown into other opcode fields, wreaking havoc.
To trigger these problems, you can try running the "movi d8, #0x0" AArch64
instruction on 32-bit x86. "vpxor %xmm0, %xmm0, %xmm0" should be generated,
but instead TCG generated "vpxor %xmm0, %xmm0, %xmm2".
Precomputing the hash values allows us to perform more frequent
accesses to the hash table, thereby reaching higher throughputs.
We keep the old behaviour by default, since (1) we might confuse
users if they measured a speedup without changing anything in
the QHT implementation, and (2) benchmarking the hash function
"on line" is also valuable.
Accessing the HT from an iterator results almost always
in a deadlock. Given that only one qht-internal function
uses this argument, drop it from the interface.
Emilio G. Cota [Fri, 17 Aug 2018 22:15:52 +0000 (18:15 -0400)]
test-qht: speed up + test qht_resize
Perform first the tests that exercise code paths that are
easier to hit at small table sizes, and then resize the table
to speed up subsequent tests. If this resize is not too large,
we can make the test faster with no code coverage loss.
The coverage increase is entirely due to calling qht_resize,
which we weren't calling before. Note that the code paths
that remain to be tested are either error handling or
can only occur when several threads are accessing the
hash table concurrently (e.g. seqlock retry, trylock fail).
- Without gcov:
Before: 1.987s
After: 0.528s
The speedup is almost the same as with gcov, although the
"before" run is a lot faster.
Emilio G. Cota [Wed, 15 Aug 2018 19:00:48 +0000 (15:00 -0400)]
qht: add qht_iter_remove
This currently has no users, but the use case is so common that I
think we must support it.
Note that without the appended we cannot safely remove a set of
elements; a 2-step approach (i.e. qht_iter first, keep track of
the to-be-deleted elements, and then a bunch of qht_remove calls)
would be racy, since between the iteration and the removals other
threads might insert additional elements.
Remove -sandbox option if the host is not capable of TSYNC, since the
sandbox will fail at setup time otherwise. This will help libvirt, for
ex, to figure out if -sandbox will work.
tests/migration: Support cross compilation in generating boot header file
Recently a new configure option, CROSS_CC_GUEST, was added to
$(TARGET)-softmmu/config-target.mak to support TCG-related tests. This
patch tries to leverage this option to support cross compilation when the
migration boot block file is being re-generated:
* The x86 related files are moved to a new sub-dir (named ./i386).
* A new top-layer Makefile is created in tests/migration/ directory.
This Makefile searches and parses CROSS_CC_GUEST to generate CROSS_PREFIX.
The CROSS_PREFIX, if available, is then passed to migration/$ARCH/Makefile.
tests/migration: Convert x86 boot block compilation script into Makefile
The x86 boot block header currently is generated with a shell script.
To better support other CPUs (e.g. aarch64), we convert the script
into Makefile. This allows us to 1) support cross-compilation easily,
and 2) avoid creating a script file for every architecture.
Note that, in the new design, the cross compiler prefix can be specified by
setting the CROSS_PREFIX in "make" command. Also to allow gcc pre-processor
to include the C-style file correctly, it also renames the
x86-a-b-bootblock.s file extension from .s to .S.
Currently, it includes:
pages: amount of pages compressed and transferred to the target VM
busy: amount of count that no free thread to compress data
busy-rate: rate of thread busy
compressed-size: amount of bytes after compression
compression-rate: rate of compressed size
migration: do not flush_compressed_data at the end of iteration
flush_compressed_data() needs to wait all compression threads to
finish their work, after that all threads are free until the
migration feeds new request to them, reducing its call can improve
the throughput and use CPU resource more effectively
We do not need to flush all threads at the end of iteration, the
data can be kept locally until the memory block is changed or
memory migration starts over in that case we will meet a dirtied
page which may still exists in compression threads's ring
This patch adds a small hint for the failure case of the load snapshot
process. It may be useful for users to remember that the VM
configuration has changed between the save and load processes.
(qemu) loadvm vm-20180903083641
Unknown savevm section or instance 'cpu_common' 4.
Make sure that your current VM setup matches your saved VM setup, including any hotplugged devices
Error -22 while loading VM state
(qemu) device_add host-spapr-cpu-core,core-id=4
(qemu) loadvm vm-20180903083641
(qemu) c
(qemu) info status
VM status: running
It also exits Qemu if the snapshot cannot be loaded before reaching the
main loop (-loadvm in the command line).
$ qemu-system-ppc64 ... -loadvm vm-20180903083641
qemu-system-ppc64: Unknown savevm section or instance 'cpu_common' 4.
Make sure that your current VM setup matches your saved VM setup, including any hotplugged devices
qemu-system-ppc64: Error -22 while loading VM state
$
As Peter pointed out:
| - xbzrle_counters.cache_miss is done in save_xbzrle_page(), so it's
| per-guest-page granularity
|
| - RAMState.iterations is done for each ram_find_and_save_block(), so
| it's per-host-page granularity
|
| An example is that when we migrate a 2M huge page in the guest, we
| will only increase the RAMState.iterations by 1 (since
| ram_find_and_save_block() will be called once), but we might increase
| xbzrle_counters.cache_miss for 2M/4K=512 times (we'll call
| save_xbzrle_page() that many times) if all the pages got cache miss.
| Then IMHO the cache miss rate will be 512/1=51200% (while it should
| actually be just 100% cache miss).
And he also suggested as xbzrle_counters.cache_miss_rate is the only
user of rs->iterations we can adapt it to count target guest page
numbers
After that, rename 'iterations' to 'target_page_count' to better reflect
its meaning
Clang correctly errors out moaning that rdma_return_path
is used uninitialised in the earlier error paths.
Make it NULL so that the error path ignores it.
There is a rare case which the size of last compressed cluster
is larger than the cluster size, which will cause the file is
not aligned at the sector boundary.
There are three reasons to do it. First, if vmdk doesn't align at
the sector boundary, there may be many undefined behaviors,
such as, in vbox it will show VMDK: Compressed image is corrupted
'syno-vm-disk1.vmdk' (VERR_ZIP_CORRUPTED) when we try to import an
ova with unaligned vmdk. Second, all the cluster_sector is aligned
to sector, the last one should be like this, too. Third, it ease
reading with sector based I/Os.
Paolo Bonzini [Wed, 12 Sep 2018 17:10:40 +0000 (19:10 +0200)]
aio-posix: do skip system call if ctx->notifier polling succeeds
Commit 70232b5253 ("aio-posix: Don't count ctx->notifier as progress when
2018-08-15), by not reporting progress, causes aio_poll to execute the
system call when polling succeeds because of ctx->notifier. This introduces
latency before the call to aio_bh_poll() and negates the advantages of
polling, unfortunately.
The fix builds on the previous patch, separating the effect of polling on
the timeout from the progress reported to aio_poll(). ctx->notifier
does zero the timeout, causing the caller to skip the system call,
but it does not report progress, so that the bug fix of commit 70232b5253
still stands.
Paolo Bonzini [Wed, 12 Sep 2018 17:10:39 +0000 (19:10 +0200)]
aio-posix: compute timeout before polling
This is a preparation for the next patch, and also a very small
optimization. Compute the timeout only once, before invoking
try_poll_mode, and adjust it in run_poll_handlers. The adjustment
is the polling time when polling fails, or zero (non-blocking) if
polling succeeds.
Paolo Bonzini [Wed, 12 Sep 2018 17:10:38 +0000 (19:10 +0200)]
aio-posix: fix concurrent access to poll_disable_cnt
It is valid for an aio_set_fd_handler to happen concurrently with
aio_poll. In that case, poll_disable_cnt can change under the heels
of aio_poll, and the assertion on poll_disable_cnt can fail in
run_poll_handlers.
Therefore, this patch simply checks the counter on every polling
iteration. There are no particular needs for ordering, since the
polling loop is terminated anyway by aio_notify at the end of
aio_set_fd_handler.
Max Filippov [Mon, 17 Sep 2018 18:13:14 +0000 (11:13 -0700)]
linux-user: do setrlimit selectively
setrlimit guest calls that affect memory resources
(RLIMIT_{AS,DATA,STACK}) may interfere with QEMU internal memory
management. They may result in QEMU lockup because mprotect call in
page_unprotect would fail with ENOMEM error code, causing infinite loop
of SIGSEGV. E.g. it happens when running libstdc++ testsuite for xtensa
target on x86_64 host.
Don't call host setrlimit for memory-related resources.
linux-user: write(fd, NULL, 0) parity with linux's treatment of same
Bring linux-user write(2) handling into line with linux for the case
of a 0-byte write with a NULL buffer. Based on a patch originally
written by Zhuowei Zhang.
Linux returns success for the special case of calling write with a
zero-length NULL buffer: compiling and running
int main() {
ssize_t ret = write(STDOUT_FILENO, NULL, 0);
fprintf(stderr, "write returned %ld\n", ret);
return 0;
}
gives "write returned 0" when run directly, but "write returned
-1" in QEMU.
This commit checks for this situation and returns success if
found.
Subsequent discussion raised the following questions (and my answers):
- Q. Should TARGET_NR_read pass through to safe_read in this
situation too?
A. I'm wary of changing unrelated code to the specific problem I'm
addressing. TARGET_NR_read is already consistent with Linux for
this case.
- Q. Do pread64/pwrite64 need to be changed similarly?
A. Experiment suggests not: both linux and linux-user yield -1 for
NULL 0-length reads/writes.
linux-user: elf: mmap all the target-pages of hostpage for data segment
If the hostpage size is greater than the TARGET_PAGESIZE, the
target-pages of size TARGET_PAGESIZE are marked valid only till the
length requested during the elfload. The glibc attempts to consume unused
space in the last page of data segment(__libc_memalign() in
elf/dl-minimal.c). If PT_LOAD p_align is greater than or
equal to hostpage size, the GLRO(dl_pagesize) is actually the host pagesize
as set in the auxillary vectors. So, there is no explicit mmap request for
the remaining target-pages on the last hostpage. The glibc assumes that
particular space as available and subsequent attempts to use
those addresses lead to crash as the target_mmap has not marked them valid
for those target-pages.
The issue is seen when trying to chroot to 16.04-x86_64 ubuntu on a PPC64
host where the fork fails to access the thread_id as it is allocated on a
page not marked valid. The recent glibc doesn't have checks for thread-id in
fork, but the issue can manifest somewhere else, none the less.
The fix here is to map all the target-pages of the hostpage during the
elfload if the p_align is greater than or equal to hostpage size, for
data segment to allow the glibc for proper consumption.
Original implementation for setsockopt by Chen Gang[1]; all bugs mine,
including removing assignment for optname which hopefully makes the
logic easier to follow and moving some variables to make the code
more selfcontained.
Peter Maydell [Tue, 25 Sep 2018 17:09:52 +0000 (18:09 +0100)]
Merge remote-tracking branch 'remotes/huth-gitlab/tags/pull-request-2018-09-25' into staging
- Deprecate the usage of a network backend via "name" instead of "id"
- Deprecate the "enforce-config-section" machine parameter
- Re-enable the wdt_ib700, endianness and vmxnet3 qtests
- Some trivial fixes and doc update patches that crossed my way
* remotes/huth-gitlab/tags/pull-request-2018-09-25:
Revert "check: Move VMXNET3 test to common"
Revert "check: Move endianess test to common"
Revert "check: Move wdt_ib700 test to common"
tests/migration: Speed up the test on ppc64
hw/qdev-core: Fix description of instance_init
qdev: fix a typo in comment
docs: Fix some typos (most found by codespell)
trivial: Make bios files and source files non-executable
memfd: fix possible usage of the uninitialized file descriptor
hw/core/machine: Officially deprecate the enforce-config-section parameter
net/slirp: Deprecate the [hub_id name] parameter tuple
net: Deprecate the "name" parameter of -net
Makefile: Add missing dependency for qemu-deprecated.texi
Peter Maydell [Tue, 25 Sep 2018 15:47:35 +0000 (16:47 +0100)]
Merge remote-tracking branch 'remotes/xanclic/tags/pull-block-2018-09-25' into staging
Block layer patches:
- Drain fixes
- node-name parameters for block-commit
- Refactor block jobs to use transactional callbacks for exiting
# gpg: Signature made Tue 25 Sep 2018 16:12:44 BST
# gpg: using RSA key F407DB0061D5CF40
# gpg: Good signature from "Max Reitz <[email protected]>"
# Primary key fingerprint: 91BE B60A 30DB 3E88 57D1 1829 F407 DB00 61D5 CF40
* remotes/xanclic/tags/pull-block-2018-09-25: (42 commits)
test-bdrv-drain: Test draining job source child and parent
block: Use a single global AioWait
test-bdrv-drain: Fix outdated comments
test-bdrv-drain: AIO_WAIT_WHILE() in job .commit/.abort
job: Avoid deadlocks in job_completed_txn_abort()
test-bdrv-drain: Test nested poll in bdrv_drain_poll_top_level()
block: Remove aio_poll() in bdrv_drain_poll variants
blockjob: Lie better in child_job_drained_poll()
block-backend: Decrease in_flight only after callback
block-backend: Fix potential double blk_delete()
block-backend: Add .drained_poll callback
block: Add missing locking in bdrv_co_drain_bh_cb()
test-bdrv-drain: Test AIO_WAIT_WHILE() in completion callback
job: Use AIO_WAIT_WHILE() in job_finish_sync()
test-blockjob: Acquire AioContext around job_cancel_sync()
test-bdrv-drain: Drain with block jobs in an I/O thread
aio-wait: Increase num_waiters even in home thread
blockjob: Wake up BDS when job becomes idle
job: Fix missing locking due to mismerge
job: Fix nested aio_poll() hanging in job_txn_apply
...
Thomas Huth [Mon, 17 Sep 2018 14:24:55 +0000 (16:24 +0200)]
tests/migration: Speed up the test on ppc64
The SLOF boot process is always quite slow ... but we can speed it up
a little bit by specifying "-nodefaults" and by using the "nvramrc"
variable instead of "boot-command" (since "nvramrc" is evaluated earlier
in the SLOF boot process than "boot-command").
Thomas Huth [Mon, 10 Sep 2018 07:46:26 +0000 (09:46 +0200)]
hw/qdev-core: Fix description of instance_init
The part of the documentation of DeviceClass that talks about instance_init
is partly wrong: instance_init() functions must not abort or exit, since
the function is also called during introspection of the device already.
So if a device calls exit() during its instance_init() function, QEMU
terminates unexpectedly if somebody tries to just have a look at the
interfaces from the device with "device_add xyz,help" or with the
"device-list-properties" QOM command. This should never happen.
Dima Stepanov [Wed, 13 Jun 2018 08:19:54 +0000 (11:19 +0300)]
memfd: fix possible usage of the uninitialized file descriptor
The qemu_memfd_alloc_check() routine allocates the fd variable on stack.
This variable is initialized inside the qemu_memfd_alloc() function.
There are several cases when *fd will be left unintialized which can
lead to the unexpected close() in the qemu_memfd_free() call.
Set file descriptor to -1 before calling the qemu_memfd_alloc routine.
Thomas Huth [Thu, 20 Sep 2018 07:22:07 +0000 (09:22 +0200)]
hw/core/machine: Officially deprecate the enforce-config-section parameter
Commit 16f7244842b5135543ef068a1adafd94c6965953 added this parameter
to the documentation, including a note that it is deprecated. But it
has never been added to the "Deprecated features" appendix, which is
our official way to deprecate legacy parameters. So let's do this now.
Thomas Huth [Thu, 20 Sep 2018 08:22:27 +0000 (10:22 +0200)]
net/slirp: Deprecate the [hub_id name] parameter tuple
The "name" in the [hub_id name] parameter tuple is the same as a
"netdev_id" (which should be unique), so specifying the hub_id here
is just redundant (it was likely just necessary in the past when
the network subsystem was still using "vlans" only and when it did
not use unique "id"s yet).
Thomas Huth [Thu, 20 Sep 2018 08:14:08 +0000 (10:14 +0200)]
net: Deprecate the "name" parameter of -net
In early times, network backends were specified by a "vlan" and "name"
tuple. With the introduction of netdevs, the "name" was replaced by an
"id" (which is supposed to be unique), but the "name" parameter stayed
as an alias which could be used instead of "id". Unfortunately, we miss
the duplication check for "name":
... starts without an error, while "id" correctly complains:
$ qemu-system-x86_64 -net user,id=n1 -net user,id=n1
qemu-system-x86_64: -net user,id=n1: Duplicate ID 'n1' for net
Instead of trying to fix the code for the legacy "name" parameter, let's
rather get rid of this old interface and deprecate the "name" parameter
now - this will also be less confusing for the users in the long run.
Peter Maydell [Tue, 25 Sep 2018 14:49:25 +0000 (15:49 +0100)]
Merge remote-tracking branch 'remotes/dgilbert/tags/pull-hmp-20180925' into staging
HMP pull 2018-09-25
# gpg: Signature made Tue 25 Sep 2018 15:11:09 BST
# gpg: using RSA key 0516331EBC5BFDE7
# gpg: Good signature from "Dr. David Alan Gilbert (RH2) <[email protected]>"
# Primary key fingerprint: 45F5 C71B 4A0C B7FB 977A 9FA9 0516 331E BC5B FDE7
* remotes/dgilbert/tags/pull-hmp-20180925:
qmp, hmp: add PCI subsystem id and vendor id to PCI info
hmp: fix migrate status timer leak
monitor: print message when using 'help' with an unknown command
Peter Maydell [Tue, 25 Sep 2018 14:24:04 +0000 (15:24 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20180925-1' into staging
target-arm queue:
* target/arm: Fix cpu_get_tb_cpu_state() for non-SVE CPUs
* hw/arm/exynos4210: fix Exynos4210 UART support
* hw/arm/virt-acpi-build: Add a check for memory-less NUMA nodes
* arm: Add BBC micro:bit machine
* aspeed/i2c: Fix interrupt handling bugs
* hw/arm/smmu-common: Fix the name of the iommu memory regions
* hw/arm/smmuv3: fix eventq recording and IRQ triggerring
* hw/intc/arm_gic: Document QEMU interface
* hw/intc/arm_gic: Drop GIC_BASE_IRQ macro
* hw/net/pcnet-pci: Convert away from old_mmio accessors
* hw/timer/cmsdk-apb-dualtimer: Add missing 'break' statements
* aspeed/timer: fix compile breakage with clang 3.4.2
* hw/arm/aspeed: change the FMC flash model of the AST2500 evb
* hw/arm/aspeed: Minor code cleanups
* target/arm: Start AArch32 CPUs with EL2 but not EL3 in Hyp mode
* remotes/pmaydell/tags/pull-target-arm-20180925-1: (21 commits)
target/arm: Start AArch32 CPUs with EL2 but not EL3 in Hyp mode
aspeed/smc: fix some alignment issues
hw/arm/aspeed: Add an Aspeed machine class
hw/arm/aspeed: change the FMC flash model of the AST2500 evb
aspeed/timer: fix compile breakage with clang 3.4.2
hw/timer/cmsdk-apb-dualtimer: Add missing 'break' statements
hw/net/pcnet-pci: Unify pcnet_ioport_read/write and pcnet_mmio_read/write
hw/net/pcnet-pci: Convert away from old_mmio accessors
hw/intc/arm_gic: Drop GIC_BASE_IRQ macro
hw/intc/arm_gic: Document QEMU interface
hw/arm/smmuv3: fix eventq recording and IRQ triggerring
hw/arm/smmu-common: Fix the name of the iommu memory regions
aspeed/i2c: Fix receive done interrupt handling
aspeed/i2c: Handle receive command in separate function
aspeed/i2c: interrupts should be cleared by software only
arm: Add BBC micro:bit machine
arm: Add Nordic Semiconductor nRF51 SoC
MAINTAINERS: Add NRF51 entry
hw/arm/virt-acpi-build: Add a check for memory-less NUMA nodes
hw/arm/exynos4210: fix Exynos4210 UART support
...
Peter Maydell [Tue, 25 Sep 2018 13:02:33 +0000 (14:02 +0100)]
target/arm: Start AArch32 CPUs with EL2 but not EL3 in Hyp mode
The ARMv8 architecture defines that an AArch32 CPU starts
in SVC mode, unless EL2 is the highest available EL, in
which case it starts in Hyp mode. (In ARMv7 a CPU with EL2
but not EL3 was not a valid configuration, but we don't
specifically reject this if the user asks for one.)
hw/arm/aspeed: change the FMC flash model of the AST2500 evb
The AST2500 evb is shipped with a W25Q256 which has a non volatile bit
to make the chip operate in 4 Byte address mode at power up. This
should be an interesting feature to model as it will exercise a bit
more the SMC controllers and MMIO execution at boot time.
aspeed/timer: fix compile breakage with clang 3.4.2
In file included from /home/thuth/devel/qemu/hw/timer/aspeed_timer.c:16:
/home/thuth/devel/qemu/include/hw/misc/aspeed_scu.h:37:3: error:
redefinition of typedef 'AspeedSCUState' is a C11 feature
[-Werror,-Wtypedef-redefinition]
} AspeedSCUState;
^
/home/thuth/devel/qemu/include/hw/timer/aspeed_timer.h:27:31: note:
previous definition is here
typedef struct AspeedSCUState AspeedSCUState;
Peter Maydell [Tue, 25 Sep 2018 13:02:32 +0000 (14:02 +0100)]
hw/net/pcnet-pci: Unify pcnet_ioport_read/write and pcnet_mmio_read/write
The only difference between our implementation of the pcnet ioport
accessors and the mmio accessors is that the former check BCR_DWIO to
see what access widths are permitted for addresses in the aprom range
(0x0..0xf). In fact our failure to do this in the mmio accessors
is a bug (one which was fixed for the ioport accessors in
commit 7ba79741970 in 2011).
The data sheet for the Am79C970A does not describe the DWIO
bit as only applying for I/O space mapped I/O resources and
not memory mapped I/O resources, and our MMIO accessors already
honour DWIO for accesses in the 0x10..0x1f range (since the
pcnet_ioport_{read,write}{w,l} functions check it).
The data sheet for the later but compatible Am79C976 is clearer:
it states specifically "DWIO mode applies to both I/O- and
memory-mapped acceses." This seems to be reasonable evidence
in favour of interpretating the Am79C970A spec as being the same.
(NB: Linux's pcnet driver only supports I/O accesses, so the
MMIO access part of this device is probably untested anyway.)
Peter Maydell [Tue, 25 Sep 2018 13:02:32 +0000 (14:02 +0100)]
hw/net/pcnet-pci: Convert away from old_mmio accessors
Convert the pcnet-pci device away from using the old_mmio
MemoryRegionOps accessor functions.
This commit is a no-behaviour-change API conversion.
(Since PCNET_PNPMMIO_SIZE is 0x20, the old "addr & 0x10"
check and the new "addr < 0x10" check are exact opposites;
the new code is phrased to be parallel with the
pcnet_io_read/write functions.)
I have left a TODO comment marker because the similarity
between the MMIO and IO accessor behaviour is suspicious
and they could be combined, but this will be left to a
different patch.
Peter Maydell [Tue, 25 Sep 2018 13:02:32 +0000 (14:02 +0100)]
hw/intc/arm_gic: Drop GIC_BASE_IRQ macro
The GIC_BASE_IRQ macro is a leftover from when we shared code
between the GICv2 and the v7M NVIC. Since the NVIC is now
split off, GIC_BASE_IRQ is always 0, and we can just delete it.
Peter Maydell [Tue, 25 Sep 2018 13:02:32 +0000 (14:02 +0100)]
hw/intc/arm_gic: Document QEMU interface
The GICv2's QEMU interface (sysbus MMIO regions, IRQs,
etc) is now quite complicated with the addition of the
virtualization extensions. Add a comment in the header
file which documents it.
Eric Auger [Tue, 25 Sep 2018 13:02:32 +0000 (14:02 +0100)]
hw/arm/smmuv3: fix eventq recording and IRQ triggerring
The event queue management is broken today. Event records
are not properly written as EVT_SET_* macro was not updating
the actual event record. Also the event queue interrupt
is not correctly triggered.
Max Reitz [Tue, 25 Sep 2018 14:12:41 +0000 (16:12 +0200)]
Merge remote-tracking branch 'kevin/tags/for-upstream' into block
Block layer patches:
- Fix some jobs/drain/aio_poll related hangs
- commit: Add top-node/base-node options
- linux-aio: Fix locking for qemu_laio_process_completions()
- Fix use after free error in bdrv_open_inherit
# gpg: Signature made Tue Sep 25 15:54:01 2018 CEST
# gpg: using RSA key 7F09B272C88F2FD6
# gpg: Good signature from "Kevin Wolf <[email protected]>"
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74 56FE 7F09 B272 C88F 2FD6
* kevin/tags/for-upstream: (26 commits)
test-bdrv-drain: Test draining job source child and parent
block: Use a single global AioWait
test-bdrv-drain: Fix outdated comments
test-bdrv-drain: AIO_WAIT_WHILE() in job .commit/.abort
job: Avoid deadlocks in job_completed_txn_abort()
test-bdrv-drain: Test nested poll in bdrv_drain_poll_top_level()
block: Remove aio_poll() in bdrv_drain_poll variants
blockjob: Lie better in child_job_drained_poll()
block-backend: Decrease in_flight only after callback
block-backend: Fix potential double blk_delete()
block-backend: Add .drained_poll callback
block: Add missing locking in bdrv_co_drain_bh_cb()
test-bdrv-drain: Test AIO_WAIT_WHILE() in completion callback
job: Use AIO_WAIT_WHILE() in job_finish_sync()
test-blockjob: Acquire AioContext around job_cancel_sync()
test-bdrv-drain: Drain with block jobs in an I/O thread
aio-wait: Increase num_waiters even in home thread
blockjob: Wake up BDS when job becomes idle
job: Fix missing locking due to mismerge
job: Fix nested aio_poll() hanging in job_txn_apply
...
Kevin Wolf [Thu, 20 Sep 2018 15:39:13 +0000 (17:39 +0200)]
test-bdrv-drain: Test draining job source child and parent
For the block job drain test, don't only test draining the source and
the target node, but create a backing chain for the source
(source_backing <- source <- source_overlay) and test draining each of
the nodes in it.
When using iothreads, the source node (and therefore the job) is in a
different AioContext than the drain, which happens from the main
thread. This way, the main thread waits in AIO_WAIT_WHILE() for the
iothread to make process and aio_wait_kick() is required to notify it.
The test validates that calling bdrv_wakeup() for a child or a parent
node will actually notify AIO_WAIT_WHILE() instead of letting it hang.
Increase the sleep time a bit (to 1 ms) because the test case is racy
and with the shorter sleep, it didn't reproduce the bug it is supposed
to test for me under 'rr record -n'.
This was because bdrv_drain_invoke_entry() (in the main thread) was only
called after the job had already reached the pause point, so we got a
bdrv_dec_in_flight() from the main thread and the additional
aio_wait_kick() when the job becomes idle (that we really wanted to test
here) wasn't even necessary any more to make progress.
Kevin Wolf [Tue, 18 Sep 2018 15:09:16 +0000 (17:09 +0200)]
block: Use a single global AioWait
When draining a block node, we recurse to its parent and for subtree
drains also to its children. A single AIO_WAIT_WHILE() is then used to
wait for bdrv_drain_poll() to become true, which depends on all of the
nodes we recursed to. However, if the respective child or parent becomes
quiescent and calls bdrv_wakeup(), only the AioWait of the child/parent
is checked, while AIO_WAIT_WHILE() depends on the AioWait of the
original node.
Fix this by using a single AioWait for all callers of AIO_WAIT_WHILE().
This may mean that the draining thread gets a few more unnecessary
wakeups because an unrelated operation got completed, but we already
wake it up when something _could_ have changed rather than only if it
has certainly changed.
Apart from that, drain is a slow path anyway. In theory it would be
possible to use wakeups more selectively and still correctly, but the
gains are likely not worth the additional complexity. In fact, this
patch is a nice simplification for some places in the code.
Kevin Wolf [Thu, 20 Sep 2018 13:51:21 +0000 (15:51 +0200)]
test-bdrv-drain: Fix outdated comments
Commit 89bd030533e changed the test case from using job_sleep_ns() to
using qemu_co_sleep_ns() instead. Also, block_job_sleep_ns() became
job_sleep_ns() in commit 5d43e86e11f.
In both cases, some comments in the test case were not updated. Do that
now.
Kevin Wolf [Thu, 13 Sep 2018 12:39:14 +0000 (14:39 +0200)]
test-bdrv-drain: AIO_WAIT_WHILE() in job .commit/.abort
This adds tests for calling AIO_WAIT_WHILE() in the .commit and .abort
callbacks. Both reasons why .abort could be called for a single job are
tested: Either .run or .prepare could return an error.
Kevin Wolf [Thu, 13 Sep 2018 12:35:27 +0000 (14:35 +0200)]
job: Avoid deadlocks in job_completed_txn_abort()
Amongst others, job_finalize_single() calls the .prepare/.commit/.abort
callbacks of the individual job driver. Recently, their use was adapted
for all block jobs so that they involve code calling AIO_WAIT_WHILE()
now. Such code must be called under the AioContext lock for the
respective job, but without holding any other AioContext lock.
Kevin Wolf [Thu, 6 Sep 2018 10:28:10 +0000 (12:28 +0200)]
test-bdrv-drain: Test nested poll in bdrv_drain_poll_top_level()
This is a regression test for a deadlock that could occur in callbacks
called from the aio_poll() in bdrv_drain_poll_top_level(). The
AioContext lock wasn't released and therefore would be taken a second
time in the callback. This would cause a possible AIO_WAIT_WHILE() in
the callback to hang.
Kevin Wolf [Fri, 17 Aug 2018 12:58:49 +0000 (14:58 +0200)]
block: Remove aio_poll() in bdrv_drain_poll variants
bdrv_drain_poll_top_level() was buggy because it didn't release the
AioContext lock of the node to be drained before calling aio_poll().
This way, callbacks called by aio_poll() would possibly take the lock a
second time and run into a deadlock with a nested AIO_WAIT_WHILE() call.
However, it turns out that the aio_poll() call isn't actually needed any
more. It was introduced in commit 91af091f923, which is effectively
reverted by this patch. The cases it was supposed to fix are now covered
by bdrv_drain_poll(), which waits for block jobs to reach a quiescent
state.
Kevin Wolf [Fri, 7 Sep 2018 13:31:22 +0000 (15:31 +0200)]
blockjob: Lie better in child_job_drained_poll()
Block jobs claim in .drained_poll() that they are in a quiescent state
as soon as job->deferred_to_main_loop is true. This is obviously wrong,
they still have a completion BH to run. We only get away with this
because commit 91af091f923 added an unconditional aio_poll(false) to the
drain functions, but this is bypassing the regular drain mechanisms.
However, just removing this and telling that the job is still active
doesn't work either: The completion callbacks themselves call drain
functions (directly, or indirectly with bdrv_reopen), so they would
deadlock then.
As a better lie, tell that the job is active as long as the BH is
pending, but falsely call it quiescent from the point in the BH when the
completion callback is called. At this point, nested drain calls won't
deadlock because they ignore the job, and outer drains will wait for the
job to really reach a quiescent state because the callback is already
running.
Kevin Wolf [Thu, 6 Sep 2018 15:47:22 +0000 (17:47 +0200)]
block-backend: Decrease in_flight only after callback
Request callbacks can do pretty much anything, including operations that
will yield from the coroutine (such as draining the backend). In that
case, a decreased in_flight would be visible to other code and could
lead to a drain completing while the callback hasn't actually completed
yet.
Note that reordering these operations forbids calling drain directly
inside an AIO callback. As Paolo explains, indirectly calling it is
okay:
- Calling it through a coroutine is okay, because then
bdrv_drained_begin() goes through bdrv_co_yield_to_drain() and you
have in_flight=2 when bdrv_co_yield_to_drain() yields, then soon
in_flight=1 when the aio_co_wake() in the AIO callback completes, then
in_flight=0 after the bottom half starts.
- Calling it through a bottom half would be okay too, as long as the AIO
callback remembers to do inc_in_flight/dec_in_flight just like
bdrv_co_yield_to_drain() and bdrv_co_drain_bh_cb() do
A few more important cases that come to mind:
- A coroutine that yields because of I/O is okay, with a sequence
similar to bdrv_co_yield_to_drain().
- A coroutine that yields with no I/O pending will correctly decrease
in_flight to zero before yielding.
- Calling more AIO from the callback won't overflow the counter just
because of mutual recursion, because AIO functions always yield at
least once before invoking the callback.
Kevin Wolf [Fri, 7 Sep 2018 11:45:54 +0000 (13:45 +0200)]
block-backend: Fix potential double blk_delete()
blk_unref() first decreases the refcount of the BlockBackend and calls
blk_delete() if the refcount reaches zero. Requests can still be in
flight at this point, they are only drained during blk_delete():
At this point, arbitrary callbacks can run. If any callback takes a
temporary BlockBackend reference, it will first increase the refcount to
1 and then decrease it to 0 again, triggering another blk_delete(). This
will cause a use-after-free crash in the outer blk_delete().
Fix it by draining the BlockBackend before decreasing to refcount to 0.
Assert in blk_ref() that it never takes the first refcount (which would
mean that the BlockBackend is already being deleted).
Kevin Wolf [Thu, 6 Sep 2018 15:43:49 +0000 (17:43 +0200)]
block-backend: Add .drained_poll callback
A bdrv_drain operation must ensure that all parents are quiesced, this
includes BlockBackends. Otherwise, callbacks called by requests that are
completed on the BDS layer, but not quite yet on the BlockBackend layer
could still create new requests.
projects / qemu.git / log