t;
> are available in the Git repository at:
>
> https://github.com/philmd/qemu.git tags/memory-api-20220118
>
> for you to fetch changes up to 9d696cd50442327fd71ec7309e7b0c6fee693b1d:
>
> docs/devel: add some clarifying text
18.01.2022 19:37, Hanna Reitz wrote:
On 22.12.21 18:40, Vladimir Sementsov-Ogievskiy wrote:
FleecingState represents state shared between copy-before-write filter
and upcoming fleecing block driver.
Signed-off-by: Vladimir Sementsov-Ogievskiy
---
block/fleecing.h | 135
On Tue, Jan 18, 2022 at 7:13 AM Peter Maydell wrote:
>
> On Mon, 17 Jan 2022 at 20:35, John Snow wrote:
> > I do expect this to print more information on failure than it
> > currently is, though (bug somewhere in machine.py, I think).
> > Can you please try applying this temporary patch and
On Tue, Jan 18, 2022 at 7:12 PM Hanna Reitz wrote:
>
> Add a new test to verify that want_zero=false block-status calls do not
> pollute the block-status cache for want_zero=true calls.
>
> We check want_zero=true calls and their results using `qemu-img map`
> (over NBD), and want_zero=false
We update the block-status cache whenever we get new information from a
bdrv_co_block_status() call to the block driver. However, if we have
passed want_zero=false to that call, it may flag areas containing zeroes
as data, and so we would update the block-status cache with wrong
information.
Add a new test to verify that want_zero=false block-status calls do not
pollute the block-status cache for want_zero=true calls.
We check want_zero=true calls and their results using `qemu-img map`
(over NBD), and want_zero=false calls also using `qemu-img map` over
NBD, but using the
On 17.01.22 18:57, Nir Soffer wrote:
On Mon, Jan 17, 2022 at 6:26 PM Hanna Reitz wrote:
Add a new test to verify that want_zero=false block-status calls do not
pollute the block-status cache for want_zero=true calls.
We check want_zero=true calls and their results using `qemu-img map`
(over
Same as the locked version, but use BDRV_POLL_UNLOCKED
Signed-off-by: Emanuele Giuseppe Esposito
---
block/io.c | 50 +---
include/block/block-io.h | 2 ++
2 files changed, 39 insertions(+), 13 deletions(-)
diff --git a/block/io.c b/block/io.c
Hi,
Cover letter from v1:
https://lists.nongnu.org/archive/html/qemu-block/2022-01/msg00314.html
In v2, I’ve added a comment to the new test explaining why we need to
pass --allocation-depth to qemu-nbd, as requested by Nir; and as he has
implied, using the short options is kind of silly because
Using bdrv_do_drained_begin_quiesce() in bdrv_child_cb_drained_end()
is not a good idea: the callback might be called when running
a drain in a coroutine, and bdrv_drained_begin_poll() does not
handle that case, resulting in assertion failure.
Instead, bdrv_do_drained_begin with no recursion and
As explained in the TODO, complete the function by checking
that the node is also drained.
In this way, we can ensure that modify the bs is thread safe,
as the drain makes sure that no I/O concurrently reads the field,
and all writes are under BQL.
Signed-off-by: Emanuele Giuseppe Esposito
---
Doing the opposite can make adding the child node to a non-drained node,
as apply_subtree_drain is only done in ->attach() and thus make
assert_bdrv_graph_writable fail.
This can happen for example during a transaction rollback (test 245,
test_io_with_graph_changes):
1. a node is removed from the
Doing the opposite can make ->detach() (more precisely
bdrv_unapply_subtree_drain() in bdrv_child_cb_detach) undo the subtree_drain
just performed to protect the removal of the child from the graph,
thus making the fully-enabled assert_bdrv_graph_writable fail.
Note that
This test uses a callback of an I/O function (blk_aio_preadv)
to modify the graph, using bdrv_attach_child.
This is simply not allowed anymore. I/O cannot change the graph.
Before "block/io.c: make bdrv_do_drained_begin_quiesce static
and introduce bdrv_drained_begin_no_poll", the test would
Add some missing assertion in static functions of block.c
Signed-off-by: Emanuele Giuseppe Esposito
---
block.c | 2 ++
block/block-backend.c | 3 +++
2 files changed, 5 insertions(+)
diff --git a/block.c b/block.c
index 6196c95aae..7961f5a984 100644
--- a/block.c
+++ b/block.c
Protect bdrv_replace_child_noperm, as it modifies the
graph by adding/removing elements to .children and .parents
list of a bs. Use the newly introduced
bdrv_subtree_drained_{begin/end}_unlocked drains to achieve
that and be free from the aiocontext lock.
One important criteria to keep in mind is
Depending on the options given to reopen_state,
bdrv_reopen_parse_file_or_backing could pick another bs
that could be from another graph, and thus not protected
by subtree_drained_begin called by the callers of this
function.
We can't simply drain-undrain here, because of transactions.
To
On 22.12.21 18:40, Vladimir Sementsov-Ogievskiy wrote:
FleecingState represents state shared between copy-before-write filter
and upcoming fleecing block driver.
Signed-off-by: Vladimir Sementsov-Ogievskiy
---
block/fleecing.h | 135 ++
block/fleecing.c |
There will be 2 problems in this test when we will add
subtree drains in bdrv_replace_child_noperm:
- First of all, inconsistency between block_job_create under
aiocontext lock that internally calls blk_insert_bs and therefore
bdrv_replace_child_noperm, and blk_insert_bs that is called two lines
If a drain happens while a job is sleeping, the timeout
gets cancelled and the job continues once the drain ends.
This is especially bad for the sleep performed in commit and stream
jobs, since that is dictated by ratelimit to maintain a certain speed.
Basically the execution path is the
This serie aims to ensure necessary protection to the BlockDriverState
.children and .parents lists, modified in block.c:bdrv_replace_child_noperm().
Thanks to the assertion qemu_in_main_thread() introduced in "block layer:
split block APIs in global state and I/O", we can verify that these lists
Same as BDRV_POLL_WHILE, but uses AIO_WAIT_WHILE_UNLOCKED.
Signed-off-by: Emanuele Giuseppe Esposito
---
include/block/block-global-state.h | 5 +
1 file changed, 5 insertions(+)
diff --git a/include/block/block-global-state.h
b/include/block/block-global-state.h
index
On 1/18/22 12:08 AM, John Snow wrote:
On Mon, Jan 17, 2022 at 3:49 PM Peter Maydell wrote:
On Mon, 17 Jan 2022 at 20:35, John Snow wrote:
Can you please try applying this temporary patch and running `./check
-qcow2 040 041` until you see a breakage and show me the output from
that?
With
On 22.12.21 18:40, Vladimir Sementsov-Ogievskiy wrote:
Add a convenient function similar with bdrv_block_status() to get
status of dirty bitmap.
Signed-off-by: Vladimir Sementsov-Ogievskiy
---
include/block/dirty-bitmap.h | 2 ++
include/qemu/hbitmap.h | 11 +++
On 22.12.21 18:40, Vladimir Sementsov-Ogievskiy wrote:
Add function to wait for all intersecting requests.
To be used in the further commit.
Signed-off-by: Vladimir Sementsov-Ogievskiy
---
include/block/reqlist.h | 8
block/reqlist.c | 8
2 files changed, 16
On 15/12/2021 21.56, Philippe Mathieu-Daudé wrote:
Include the qtest reproducer provided by Alexander Bulekov
in https://gitlab.com/qemu-project/qemu/-/issues/451. Without
the previous commit, we get:
$ make check-qtest-i386
...
Running test qtest-i386/fuzz-sdcard-test
Reviewed-by: Richard Henderson
Reviewed-by: David Hildenbrand
Signed-off-by: Philippe Mathieu-Daudé
Message-Id: <2022084309.28637-10-f4...@amsat.org>
---
hw/ide/ahci.c| 2 +-
hw/rdma/trace-events | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/hw/ide/ahci.c
From: Philippe Mathieu-Daudé
Update the obvious places where dma_addr_t should be used
(instead of uint64_t, hwaddr, size_t, int32_t types).
This allows to have _addr_t type portable on 32/64-bit
hosts.
Signed-off-by: Philippe Mathieu-Daudé
Reviewed-by: David Hildenbrand
Signed-off-by:
From: Philippe Mathieu-Daudé
DMA API should not be included in user-mode emulation.
If so, build should fail. Remove the CONFIG_USER_ONLY check.
Signed-off-by: Philippe Mathieu-Daudé
Reviewed-by: Richard Henderson
Reviewed-by: David Hildenbrand
Message-Id:
From: Alex Bennée
We do mention the limitation of single parenthood for
memory_region_add_subregion but lets also make it clear how aliases
help solve that conundrum.
Signed-off-by: Alex Bennée
Reviewed-by: Philippe Mathieu-Daudé
Message-Id: <20220110175104.2908956-7-alex.ben...@linaro.org>
On Mon, 17 Jan 2022 at 20:35, John Snow wrote:
> I do expect this to print more information on failure than it
> currently is, though (bug somewhere in machine.py, I think).
> Can you please try applying this temporary patch and running `./check
> -qcow2 040 041` until you see a breakage and show
In the next commit we will use the dma_addr_t type in the QEMUSGList
structure. Since currently dma_addr_t is defined after QEMUSGList,
move the declarations to have dma_addr_t defined first. This is a
pure code-movement patch.
Suggested-by: David Hildenbrand
Signed-off-by: Philippe
From: David Hildenbrand
memory_region_is_mapped() currently does not return "true" when a memory
region is mapped via an alias.
Assuming we have:
alias (A0) -> alias (A1) -> region (R0)
Mapping A0 would currently only make memory_region_is_mapped() succeed
on A0, but not on A1 and R0.
From: Philippe Mathieu-Daudé
Signed-off-by: Philippe Mathieu-Daudé
Reviewed-by: Michael S. Tsirkin
Reviewed-by: Richard Henderson
Message-Id: <2022084309.28637-5-f4...@amsat.org>
Signed-off-by: Philippe Mathieu-Daudé
---
include/hw/pci/pci.h | 12
1 file changed, 12
From: Philippe Mathieu-Daudé
Neither tools nor user-mode emulation require the PCI bus stub.
Signed-off-by: Philippe Mathieu-Daudé
Reviewed-by: Michael S. Tsirkin
Reviewed-by: Richard Henderson
Reviewed-by: David Hildenbrand
Message-Id: <2022084309.28637-4-f4...@amsat.org>
From: Peter Xu
We should only call the log_global_start/stop when the global dirty track
bitmask changes from zero<->non-zero.
No real issue reported for this yet probably because no immediate user to
enable both dirty rate measurement and migration at the same time. However
it'll be good to
From: Philippe Mathieu-Daudé
Various APIs use 'pval' naming for 'pointer to val'.
rdma_pci_dma_map() uses 'plen' for 'PCI length', but since
'PCI' is already explicit in the function name, simplify
and rename the argument 'len'. No logical change.
Signed-off-by: Philippe Mathieu-Daudé
From: Philippe Mathieu-Daudé
fw_cfg QOM interface is required by system emulation and
qemu-storage-daemon. User-mode emulation doesn't need it.
Signed-off-by: Philippe Mathieu-Daudé
Signed-off-by: Philippe Mathieu-Daudé
Reviewed-by: David Hildenbrand
Message-Id:
From: Philippe Mathieu-Daudé
The 'resid' field is slightly confusing and could be
interpreted as some ID. Rename it as 'residual' which
is clearer to review. No logical change.
Signed-off-by: Philippe Mathieu-Daudé
Reviewed-by: Richard Henderson
Reviewed-by: David Hildenbrand
Message-Id:
fw_cfg_arch_key_name() stub is only required for sysemu.
Signed-off-by: Philippe Mathieu-Daudé
Reviewed-by: David Hildenbrand
Message-Id: <2022084309.28637-2-f4...@amsat.org>
---
stubs/meson.build | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/stubs/meson.build
From: David Hildenbrand
Let's update the documentation, making it clearer what the semantics
of memory_region_is_mapped() actually are.
Reviewed-by: Philippe Mathieu-Daudé
Reviewed-by: Peter Xu
Signed-off-by: David Hildenbrand
Message-Id: <20211102164317.45658-4-da...@redhat.com>
From: Philippe Mathieu-Daudé
Per Peter Maydell [*]:
'info mtree' monitor command was designed on the assumption that
there's really only one or two interesting address spaces, and
with more recent developments that's just not the case any more.
Similarly about how the FlatView are sorted
From: Philippe Mathieu-Daudé
While mtree_info() handles both ASes and flatviews cases,
the two cases share basically no code. Split mtree_info()
as mtree_info_flatview() + mtree_info_as() to simplify.
Suggested-by: Peter Maydell
Reviewed-by: David Hildenbrand
Reviewed-by: Peter Xu
From: Philippe Mathieu-Daudé
Since commit 292e13142d2, dma_buf_rw() returns a MemTxResult type.
Do not discard it, return it to the caller. Pass the previously
returned value (the QEMUSGList residual size, which was rarely used)
as an optional argument.
With this new API, SCSIRequest::residual
-20220118
for you to fetch changes up to 9d696cd50442327fd71ec7309e7b0c6fee693b1d:
docs/devel: add some clarifying text for aliases (2022-01-18 12:56:29 +0100)
Memory API patches
- Directly dispatch MemoryRegion alias accesses
- Remove
Since commit 2cdfcf272d ("memory: assign MemoryRegionOps to all
regions"), all newly created regions are assigned with
unassigned_mem_ops (which might be then overwritten).
When using aliased container regions, and there is no region mapped
at address 0 in the container, the
From: David Hildenbrand
memory_region_is_mapped() is the wrong check, we actually want to check
whether the backend is already marked mapped.
For example, memory regions mapped via an alias, such as NVDIMMs,
currently don't make memory_region_is_mapped() return "true". As the
machine is
On Mon, 17 Jan 2022 at 23:09, John Snow wrote:
> Well, today I learned that:
>
> (1) vm-build-XXX targets use your host system's QEMU to run that VM
> (2) my QMP library cannot talk to QEMU 2.11.
Whoops, I hadn't realised I was running with that ancient a QEMU.
For the scripted runs I set the
48 matches
Mail list logo
