es the fact that virt-resize is doing MUCH more than resizing the
block device - it is also copying entire filesystems from the old
location to the new resized location for optimal initial placement.
--
Eric Blake eblake redhat com+1-919-301-3266
Libvirt virtualization library ht
tion,
it shouldn't matter if the drive-backup failed to send all guest sectors
modified after the point in time, so long as every sector you read is
accurate (either because the guest hasn't touched it in the meantime; or
because even though the guest touched it after the point in time, y
optimizing out
the writes to the NBD target for sectors you don't care about, and
doesn't quite address the desire for making random reads take priority
over linear streaming of dirty blocks, but it might help.
--
Eric Blake eblake redhat com
mu-img. As for why qcow2 creation is
slow, I don't know what we can do about it.
>
> Any thoughts on this before I implement something ...
We definitely need to tie it into the XML that has already been
designated for this purpose: .
--
Eric Blake ebl...@redhat.com+1-919-301-32
On 10/05/2012 06:29 AM, Eric Blake wrote:
> On 10/05/2012 05:53 AM, Richard W.M. Jones wrote:
>>
>> And looking even more closely, I see disk->transient (not
>> implemented in the qemu driver ... why?)
>
> Lack of developer time :) [My usual excuse?
On 10/05/2012 05:53 AM, Richard W.M. Jones wrote:
>
> And looking even more closely, I see disk->transient (not
> implemented in the qemu driver ... why?)
Lack of developer time :) [My usual excuse?]
--
Eric Blake ebl...@redhat.com+1-919-301-3266
Libvirt virtualization
>
> The image is the 222G /home partition only. The other partitions are
> on another image.
>
> Any ideas how to troubleshoot this would be greatly appreciated.
> Thanks,
> Alex
>
> ___
> libvirt-users mailing list
&
rather than our current policy of tracking down all
contributors and asking them to use their grant-back clause of their FSF
copyright assignment as our backdoor of not having to involve the FSF.
--
Eric Blake eblake redhat com+1-919-301-3266
Libvirt virtualization library http://libvi
implementation; but in the meantime, my work
to allow filters to open independent backends has made it feasible to
implement.
Eric Blake (2):
retry: Add in retry support during .open
retry: Test previous patch
tests/Makefile.am| 2 +
server/backend.c | 7 ++-
filters/retry
Now that a filter can open a backend as many times as it wants,
there's no longer a technical reason we can't retry .open. However,
adding retry logic here does mean we have to weaken an assert in the
server backend code, since prepare can now be reached more than once.
Test coverage will be adde
Separate commit to allow rearranging patches to demonstrate that open
can now be retried.
---
tests/Makefile.am| 2 +
tests/test-retry-open.sh | 85
2 files changed, 87 insertions(+)
create mode 100755 tests/test-retry-open.sh
diff --git a/tests/
On Fri, Jan 27, 2023 at 02:41:22PM -0600, Eric Blake wrote:
> Now that a filter can open a backend as many times as it wants,
> there's no longer a technical reason we can't retry .open. However,
> adding retry logic here does mean we have to weaken an assert in the
> serv
On Sat, Jan 28, 2023 at 01:47:32PM +, Richard W.M. Jones wrote:
> On Fri, Jan 27, 2023 at 02:41:22PM -0600, Eric Blake wrote:
> > Now that a filter can open a backend as many times as it wants,
> > there's no longer a technical reason we can't retry .open. However,
&g
the name "stored"
assigned. Note that, if multiple file descriptors are submitted at
once, the specified name will be assigned to all of them. In order
to assign different names to submitted file descriptors, submit
them in separate invocations of sd_pid_notify_with_fds(). The nam
ies or by
my idea of an alternative API that takes the socket name alongside the
argv; and whether we keep to our 32-byte [[:alnum:]] limit or instead
allow for a larger name up to 255 bytes in the regex range notation by
ASCII byte value [\x20-\x39\x41-\x7e] (aka [ -9;-~], or
[^\x00-\x1f\x7f-\xff]
On Tue, Jan 31, 2023 at 11:52:27AM +, Richard W.M. Jones wrote:
> On Tue, Jan 31, 2023 at 01:33:25PM +0200, Nir Soffer wrote:
> > On Tue, Jan 31, 2023 at 12:34 AM Richard W.M. Jones
> > wrote:
> > >
> > > On Fri, Nov 04, 2022 at 04:18:31PM -0500, Eric
gt; >
> > execvp() scans PATH, and is not safe to use in this concept.
>
> That's quite annoying.
>
> > I think we should call execve() instead. First, it is async-signal-safe.
> > Second, it could take "env.ptr" directly; I do find the "environ"
&g
You have a strong point there. Just because systemd allows it doesn't
make it wise; I'm a big fan of "it's easier to resrict now and loosen
later when we see the need", as being easier than "let's try and be as
loose as we can now, then regre
guestfs-release-notes-1.50.pod#pod-errors
I presume there's a way to mark the .pod file, or at least tweak the
Makefile that renders the page from .pod, to state that the input is
intentionally UTF-8, in order to clear out that error.
--
Eric Blake, Principal Software Engineer
Red Hat, Inc.
can_multi_conn: true
> $ ./nbdkit curl file:/var/tmp/jammy-server-cloudimg-amd64.raw --run '
> nbdinfo $uri ' | grep can_multi_conn
> can_multi_conn: false
>
> This improves performance.
Reviewed-by: Eric Blake
--
Eric Blake, Principal Software Engineer
Red
ds, it seems like qemu-img convert
> above could go a lot faster if we allowed multi-conn.
>
> Is there any work on adding multi-conn support to qemu's NBD client?
Not that I'm aware of at the moment, but we have proof that it may
prove fruitful to have someone spend time on.
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
___
Libguestfs mailing list
Libguestfs@redhat.com
https://listman.redhat.com/mailman/listinfo/libguestfs
> real3m53.923s
> user0m13.751s
> sys 0m15.346s
>
> Now this is not something I'm personally concerned about (since I've
> long been arguing we should deprecate the qemu curl driver and use
> nbdkit), but
On Sun, Feb 05, 2023 at 09:10:29AM +0100, Laszlo Ersek wrote:
> On 2/1/23 17:17, Laszlo Ersek wrote:
> > On 1/31/23 18:19, Eric Blake wrote:
>
> >> The alternative to relying on execvp() to scan PATH is to pre-scan
> >> PATH ourselves before fork(). I wish t
w;
copy/file-ops.c:data = (char *) data + r;
I think I'm a fan of having the space after the cast operator, and as
long as we're doing tree-wide cleanups, this would be a good time to
inject such a patch if we wanted.
Also, it might be cool to have a code formatting tool automatically
n in this patch.
And while I mentioned in 01/29 about the possibility of a C code
formatter as a CI step for *.[ch], it would be even harder to insist
that generated code matches a given style (it's always a nice goal for
generated files to be human-readable where possible, but I'd mu
-off-by: Laszlo Ersek
> ---
> lib/internal.h | 4 ++--
> generator/states-connect-socket-activation.c | 16
> lib/handle.c | 12 ++--
> 3 files changed, 16 insertions(+), 16 deletions(-)
Reviewed-by: Eric Blake
by: Laszlo Ersek
> ---
> ocaml/helpers.c | 8
> 1 file changed, 4 insertions(+), 4 deletions(-)
Reviewed-by: Eric Blake
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
__
ector *env)
>
> err:
>set_error (errno, "malloc");
> - string_vector_iter (env, (void *) free);
> - free (env->ptr);
> + string_vector_empty (env);
>return -1;
At any rate, I like how the callers look improved, even if we aren't
quite settled on how the vector.h changes should end up.
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
___
Libguestfs mailing list
Libguestfs@redhat.com
https://listman.redhat.com/mailman/listinfo/libguestfs
al case, but it's good enough
> for writing out diagnostics before giving up.)
>
> Signed-off-by: Laszlo Ersek
> ---
> lib/utils.c | 39 ++--
> 1 file changed, 35 insertions(+), 4 deletions(-)
Reviewed-by: Eric Blake
--
Eric Blake, Principal Software En
t(). But POSIX says that abort() shall attempt to
override any SIGABRT handler in place, so I wonder if glibc's
implementation is a bit too defensive. At any rate, your
implementation looks reasonable, and more to the point, satisfies your
desire of being async-signal-safe and thus appropriat
s
would need to forward that on.
> +signal_name=$(kill -l $exit_status)
> +test "x$signal_name" = xABRT || test "x$signal_name" = xSIGABRT
> +
> +ptrn="^test-fork-safe-assert\\.c:[0-9]+: main: Assertion \`FALSE'
> failed\\.\$&quo
cvpe_init().
Makes sense. A shame that it requires 3 functions, and that we have
to duplicate so much work that libc would normally do for us if we
don't care about async-safety, but I think the result is worth it.
>
> The main idea with the above distribution of responsibilities is tha
fatal otherwise, we do get that error number.
> +run empty:fifo:nxregf:symlink f
> +execve_fail empty/f,fifo/f,nxregf/f,symlink/f ELOOP
> +
> +# Put a single prefix in PATH, such that it lead to a successful execution.
> This
leads
> +# exercises two things at the same time: (a) t
On Wed, Feb 15, 2023 at 05:03:48PM -0600, Eric Blake wrote:
> > +
> > +# Create subdirectories for triggering non-fatal internal error conditions
> > of
> > +# execvpe(). (Almost) every subdirectory will contain one entry, called
> > "f".
> > +#
>
t;malloc" would work too (we have several places
where we aren't precise on WHAT allocation failed, but where it's
obvious from the accompanying ENOMEM strerror() message that it was an
allocation failure). But when it's easy to be precise, go for it.
Reviewed-by: Eric Bl
with "N", not "O". That would be an
alternative to decreasing the refcount of py_array on success, but not
eliminate the need to decrease the refcount on Py_BuildValue failure.
>
> - when "args" is killed (decref'd), it takes
"...O", py_array);
Py_DECREF (py_array);
if (args == NULL) {
PyErr_PrintEx (0);
goto out;
}
...
out:
Py_DECREF (args);
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
On Fri, Feb 17, 2023 at 11:49:58AM +, Richard W.M. Jones wrote:
> On Wed, Feb 15, 2023 at 01:48:39PM -0600, Eric Blake wrote:
> > In my experience with GNU code (which this is not), the style I've
> > seen there is to omit () whenever possible, as in:
> >
> >
>if (prepare_socket_activation_environment (&env) == -1) {
> > SET_NEXT_STATE (%.DEAD);
> > +set_error (errno, "prepare_socket_activation_environment");
>
> Why move this out of the function?
Moving it here lets us give a more specific message
matically be enforced need not carry
baggage of a large config file. But yeah, definitely not a priority
for today.
>
> > Overall, I don't have any strong reasons to insist on shorter #ifdef
> > spellings, and the rest of your changes, while mechanical, do make the
> > codebas
On Mon, Feb 20, 2023 at 06:03:13PM +0100, Laszlo Ersek wrote:
> On 2/15/23 21:27, Eric Blake wrote:
> > On Wed, Feb 15, 2023 at 03:11:34PM +0100, Laszlo Ersek wrote:
> >> The "name##_iter" function is used 11 times in libnbd; in all those cases,
> >> "name
On Mon, Feb 20, 2023 at 07:21:05PM +0100, Laszlo Ersek wrote:
> On 2/15/23 21:57, Eric Blake wrote:
> > On Wed, Feb 15, 2023 at 03:11:36PM +0100, Laszlo Ersek wrote:
> >> Add an assert() variant that we may call between fork() and exec*().
> >>
>
On Mon, Feb 20, 2023 at 09:08:08PM +0200, Nir Soffer wrote:
> On Mon, Feb 20, 2023 at 10:45 AM Laszlo Ersek wrote:
> >
> > On 2/17/23 17:52, Eric Blake wrote:
> > > On Thu, Feb 16, 2023 at 03:09:02PM +0100, Laszlo Ersek wrote:
> >
> > >> - Py_BuildValue w
On Tue, Feb 21, 2023 at 12:55:49PM +0100, Laszlo Ersek wrote:
> On 2/15/23 23:28, Eric Blake wrote:
> > On Wed, Feb 15, 2023 at 03:11:38PM +0100, Laszlo Ersek wrote:
> >> execvp() [01] is powerful:
> >
> > You KNOW this is going to be a juicy commit message when t
On Tue, Feb 21, 2023 at 02:06:33PM +0100, Laszlo Ersek wrote:
> On 2/16/23 00:33, Eric Blake wrote:
> > On Wed, Feb 15, 2023 at 05:03:48PM -0600, Eric Blake wrote:
> >>> +
> >>> +# Create subdirectories for triggering non-fatal internal error
> >>> con
On Tue, Feb 21, 2023 at 01:53:25PM +0100, Laszlo Ersek wrote:
> On 2/16/23 00:03, Eric Blake wrote:
> > On Wed, Feb 15, 2023 at 03:11:39PM +0100, Laszlo Ersek wrote:
> >> Don't try to test async-signal-safety, but strive to exercise as many as
> >> possible paths t
ctor', and so on, per each use of
DEFINE_VECTOR_TYPE() - but while our one macro can easily declare
other type-specific functions, you can't really use the preprocessor
to define further type-specific macros. So having to tell the user to
supply their own type name when reusing the type-
On Tue, Feb 21, 2023 at 07:07:38PM +0100, Laszlo Ersek wrote:
> On 2/21/23 18:06, Eric Blake wrote:
> > On Tue, Feb 21, 2023 at 12:55:49PM +0100, Laszlo Ersek wrote:
> >> On 2/15/23 23:28, Eric Blake wrote:
> >>> On Wed, Feb 15, 2023 at 03:11:38PM +0100, Laszlo Ersek
From: Eric Blake
To: Laszlo Ersek , d...@python.org
Cc: Nir Soffer , arye...@google.com,
libguestfs@redhat.com, shtark...@google.com
Bcc:
Subject: Re: [Libguestfs] [PATCH 1/2] python: Avoid crash if callback
parameters cannot be built
Reply-To:
In-Reply-To:
[adding d...@python.org, in
VECTOR_EMPTY_METHOD() instead of DEFINE_VECTOR_EMPTY() works for
> > > me.
> >
> > OK, ADD_VECTOR_EMPTY_METHOD() can work with the above.
>
> This sounds fine for now, and since these are implementation details
> we can always revisit them in future.
--
Eric B
On Wed, Feb 22, 2023 at 01:47:02AM +0100, Laszlo Ersek wrote:
> On 2/21/23 20:32, Eric Blake wrote:
> > On Tue, Feb 21, 2023 at 07:07:38PM +0100, Laszlo Ersek wrote:
>
> >> Well, given Daniel's comments meanwhile, it seems like the original
> >> execvp() i
libnbd's copy/copy-nbd-error.sh was triggering an assertion failure in
nbdkit:
$ nbdcopy -- [ nbdkit --exit-with-parent -v --filter=error pattern 5M
error-pread-rate=0.5 ] null:
...
nbdkit: pattern.2: debug: error-inject: pread count=262144 offset=4718592
nbdkit: pattern.2: debug: pattern: pread
(exports, struct export);
static exports export_list = empty_vector;
static int
diff --git c/info/map.c i/info/map.c
index a5aad955..4924866a 100644
--- c/info/map.c
+++ i/info/map.c
@@ -36,7 +36,7 @@
#include "nbdinfo.h"
-DEFINE_VECTOR_TYPE (uint32_vector, uint32_t)
+DEFINE_VECTO
or: introduce DEFINE_POINTER_VECTOR_TYPE()
> convert string_vector_(iter(free) + reset()) to string_vector_empty()
And turned out very nicely in my opinion. See my comments on 3. For
1,2,4,5:
Reviewed-by: Eric Blake
--
Eric Blake, Principal Software Engine
On Fri, Feb 24, 2023 at 09:04:29AM -0600, Eric Blake wrote:
> On Fri, Feb 24, 2023 at 05:39:35AM +0100, Laszlo Ersek wrote:
> > A semicolon after a DEFINE_VECTOR_TYPE(...) macro invocation is not
> > needed, nor does our documentation in "common/utils/vector.h" prescribe
On Fri, Feb 24, 2023 at 06:41:49AM +0100, Laszlo Ersek wrote:
> On 2/23/23 21:40, Eric Blake wrote:
> > libnbd's copy/copy-nbd-error.sh was triggering an assertion failure in
> > nbdkit:
> >
> > $ nbdcopy -- [ nbdkit --exit-with-parent -v --filter=error pattern 5M
On Fri, Feb 24, 2023 at 01:57:37PM +0100, Laszlo Ersek wrote:
> On 2/21/23 07:33, Laszlo Ersek wrote:
> > On 2/20/23 19:21, Laszlo Ersek wrote:
> >> On 2/15/23 21:57, Eric Blake wrote:
> >>> On Wed, Feb 15, 2023 at 03:11:36PM +0100, Laszlo Ersek wrote:
> >
>
the end, I did get a working test written (by just quit()ing python
instead of trying to keep it alive). See my next email for the nbdkit
patch that spawned my request here.
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | l
libnbd's copy/copy-nbd-error.sh was triggering an assertion failure in
nbdkit (this is the command that triggered the failure, at least with
nbdcopy 1.14.2; although future nbdcopy may be fixed independently to
more gracefully exit):
$ nbdcopy -- [ nbdkit --exit-with-parent -v --filter=error patte
On Sun, Feb 26, 2023 at 10:16:08AM +0100, Laszlo Ersek wrote:
> On 2/24/23 23:59, Eric Blake wrote:
> > @@ -752,14 +742,15 @@ protocol_recv_request_send_reply (void)
> >(cmd == NBD_CMD_READ || cmd == NBD_CMD_BLOCK_STATUS)) {
> > if (!error) {
> >
olons, but that doesn't play nicely with Emacs's C
I'm not sure if the possessive is spelled "Emacs'". Either way,
Reviewed-by: Eric Blake
> language parser. Thus, force the semicolon instead.
>
> Signed-off-by: Laszlo Ersek
> ---
>
> Notes:
&g
ring_vector_empty() in a subsequent patch.
>
> Signed-off-by: Laszlo Ersek
> ---
>
> Notes:
> v5:
>
> - resolve conflict from update to previous patch
>
> - force semicolon after ADD_VECTOR_EMPTY_METHOD() and
> DEFINE_POINTER_VECTOR_TYPE() too
>
ally choose a hash that can be computed out-of-order, such
as a Merkle Tree. But we'd need a standard setup for all parties to
agree on how the hash is to be computed and checked, if it is going to
be anything more than just a linear hash of the entire guest-visible
conte
but depends on an as-yet uncommitted patch in libvirt-ci:
https://gitlab.com/libvirt/libvirt-ci/-/merge_requests/360
Eric Blake (5):
ci: Expose more env vars needed by build.sh
ci: Another attempt at MacOS
rust: Skip CI builds on MacOS
golang: Skip CI builds on MacOS and newer FreeBSD
perl:
rust is causing odd linker errors on MacOS 12:
Compiling nbdkit v0.2.0
(/private/var/folders/76/zy5ktkns50v6gt5g8r0sf6scgn/T/cirrus-ci-build/plugins/rust)
error: linking with `cc` failed: exit status: 1
|
...
= note: Undefined symbols for architecture arm64:
"_nbdkit_error", re
Import some lcitool improvements to avoid packages that are not
available through homebrew on MacOS, as well as some other tweaks:
- Bare libtorrent does not exist; libtorrent-rasterbar is available
but fails to compile due to broken pkg-config --cflags
- tcl is spelled tcl-tk
- python3-boto3, sfd
In order for build.sh to see variables on Cirrus CI, we have to make
sure they get passed through from the gitlab container. This will
make it possible to skip language bindings that don't work on FreeBSD
or MacOS.
---
.gitlab-ci.yml | 3 +++
ci/cirrus/build.yml | 3 +++
2 files changed, 6 i
perl is causing odd compiler errors, such as on FreeBSD 13:
In file included from /usr/local/lib/perl5/5.32/mach/CORE/sbox32_hash.h:4:
/usr/local/lib/perl5/5.32/mach/CORE/zaphod32_hash.h:164:5: error: '(' and '{'
tokens introducing statement expression appear in different macro expansion
context
golang is causing odd linker errors on these platforms:
FreeBSD 13:
go build -o
/tmp/cirrus-ci-build/plugins/golang/examples/disk/nbdkit-godisk-plugin.so
-buildmode=c-shared
# main
/usr/local/go119/pkg/tool/freebsd_amd64/link: running cc failed: exit status 1
ld: error: /lib/libc.so.7: undefined
On Wed, Mar 01, 2023 at 06:43:11PM +0100, Laszlo Ersek wrote:
> On 3/1/23 17:54, Eric Blake wrote:
> > I took the easy route of crippling what I couldn't get working, on the
> > grounds that partial coverage is better than none now that we have
> > Cirrus CI chec
On Fri, Mar 03, 2023 at 07:00:16PM +, Richard W.M. Jones wrote:
>
> This series looks, great, thanks!
>
> ACK
ACK from me as well; I didn't spot any problems in my read through.
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtuali
ke it
obvious that a cookie is opaque data from the point of view of the
server. Makes no difference to implementations (other than older code
still using 'handle' may be slightly harder to tie back to the spec).
Suggested-by: Richard W.M. Jones
Signed-off-by: Eric Blake
---
doc/
On Fri, Dec 16, 2022 at 10:32:01PM +0300, Vladimir Sementsov-Ogievskiy wrote:
> On 11/15/22 01:46, Eric Blake wrote:
> > The spec was silent on how many extents a server could reply with.
> > However, both qemu and nbdkit (the two server implementations known to
> >
On Fri, Dec 16, 2022 at 11:22:49PM +0300, Vladimir Sementsov-Ogievskiy wrote:
> On 11/15/22 01:46, Eric Blake wrote:
> > Commit 9f30fedb improved the spec to allow non-payload requests that
> > exceed any advertised maximum block size. Take this one step further
> > by per
st slow review time.
>
> On Mon, Nov 14, 2022 at 04:46:51PM -0600, Eric Blake wrote:
> > Commit 9f30fedb improved the spec to allow non-payload requests that
> > exceed any advertised maximum block size. Take this one step further
> > by permitting the server to use
On Wed, Feb 22, 2023 at 11:49:18AM +0200, Wouter Verhelst wrote:
> On Mon, Nov 14, 2022 at 04:46:52PM -0600, Eric Blake wrote:
> [...]
> > @@ -1370,9 +1475,10 @@ of the newstyle negotiation.
> > Return a list of `NBD_REP_META_CONTEXT` replies, one per context,
>
On Wed, Feb 22, 2023 at 12:05:44PM +0200, Wouter Verhelst wrote:
> On Mon, Nov 14, 2022 at 04:46:54PM -0600, Eric Blake wrote:
> > Simple reply message
> >
> > @@ -1232,6 +1235,19 @@ The field has the following format:
> >will be faster than a regular write
On Sun, Mar 05, 2023 at 10:53:38AM +0200, Wouter Verhelst wrote:
> On Sat, Mar 04, 2023 at 10:03:46PM +0200, Nir Soffer wrote:
> > On Sat, Mar 4, 2023 at 12:15 AM Eric Blake wrote:
> > > Makes no difference to implementations (other than older code
> > > still usi
zeof *last_error);
> if (last_error) {
>int err = pthread_setspecific (errors_key, last_error);
> - if (err != 0) {
> + /* Ignore EINVAL because that can happen in a race with other
> + * threads when we are exiting.
> + */
> + if (err
ling exit(), to work around the assertion failure in
nbdkit 1.32.5)
- various fuse/* (not sure if we can address that; cleaning up FUSE is
tricky)
This reverts a small part of commit 831cbf7ee14c ("generator: Allow
DEAD state actions to run").
Thanks: Richard W.M. Jones
Signed-off-by
On Wed, Mar 08, 2023 at 04:29:03PM -0600, Eric Blake wrote:
>
> There aren't any good ways to fix this. We have proven that the
> assertions are too tight (it IS possible for one thread's first use of
> libnbd API, which causes the allocation of a thread-local error
>
ly why libvirt uses -Z nodelete - the moment a
thread can have thread-local data with a destructor pointing to libnbd
code, but where libnbd can be removed from memory without calling
pthread_key_delete(), is the moment you get a SEGV trying to call the
thread destructor. A memory leak is
in nbd_internal_fork_safe_perror() where
> at least one of the first two write() syscalls fails, and overwrites
> "errno", before we get to formatting the error string from "errno".
All nice benefits, even if we don't normally exercise the code.
>
> Than
On Wed, Mar 15, 2023 at 03:30:12PM +0100, Laszlo Ersek wrote:
> On 3/15/23 15:01, Eric Blake wrote:
>
> > [...]
>
> Thanks for the thorough review; I'm glad all the fine points I sought to
> put in the patch were received -- and well-received! :)
>
> One qu
LENO, file, ":", line_out, ": ", func, ": Assertion `",
> + assertion, "' failed.\n", (char *)NULL);
xwritel() makes this so much shorter ;)
Reviewed-by: Eric Blake
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
xported.
On the other hand, prctl() is definitely Linux-specific, so I think
you are quite safe in assuming that exists if and only
if prctl() is a linkable entry point. If it does turn out to break
someone, we can fix it in a followup patch, so no change needed in
your usage at this time.
>
nciple -- both
> "libnbd.h" and "nbd-protocol.h" look public, at least to an extent,
> while the stuff in (3.c) is totally internal.
>
> So, I can equate wrapper macros to public headers, for now, and I won't
> introduce a new macro just for this one applica
On Thu, Mar 16, 2023 at 10:50:06AM +0100, Laszlo Ersek wrote:
> On 3/15/23 18:25, Eric Blake wrote:
> > On Wed, Mar 15, 2023 at 12:01:57PM +0100, Laszlo Ersek wrote:
> >> Don't try to test async-signal-safety, only that
> >> NBD_INTERNAL_FORK_SAFE_AS
irectory to it.
> > +tmpd=$(mktemp -d)
> > -+trap 'rm -r -- "$tmpd"' EXIT
> > ++cleanup_fn rm -r -- "$tmpd"
"$tmpd" probably starts with /, such that the -- is not strictly
necessary; but keep it (good practice, and saves u
On Tue, Mar 21, 2023 at 07:04:59AM +0100, Laszlo Ersek wrote:
> On 3/20/23 20:41, Eric Blake wrote:
> > On Sun, Mar 19, 2023 at 10:41:37AM +0100, Laszlo Ersek wrote:
> >> This is version 4 of the following sub-series:
> >>
> >> [libnbd PATCH v3 09/29] lib/utils:
for realpath deficiency on some platforms
I didn't even pay attention to this one before you pointed it out; but
it is indeed a bug in busybox now that POSIX is moving towards
standardizing realpath, so I've filed it:
https://bugs.busybox.net/show_bug.cgi?id=15466
> 2 65631e5dfff
On Wed, Mar 22, 2023 at 10:13:10AM +, Daniel P. Berrangé wrote:
> On Tue, Mar 21, 2023 at 09:05:12AM -0500, Eric Blake wrote:
> >
> > $ podman build -f ci/containers/alpine-edge.Dockerfile -t libnbd-alpine-edge
>
> You usually shouldn't need this step when rep
On Wed, Mar 22, 2023 at 03:45:17PM +0100, Laszlo Ersek wrote:
> On 3/21/23 18:28, Eric Blake wrote:
>
> > it is indeed a bug in busybox now that POSIX is moving towards
> > standardizing realpath, so I've filed it:
> > https://bugs.busybox.net/show_bug.cgi?id=15466
&g
ack when I have some URLs to list
archives of my pending questions.
>
> This can be demonstrated with:
>
> $ PATH=.:$PATH strace -etrace=execve test-execvp
>
> execve("./test-execvp", ["test-execvp"], 0x7ffc0d1e5248 /* 85
researching
for that, I found that FreeBSD had the same bug up until 2020:
https://cgit.freebsd.org/src/commit/?id=301cb491ea
On the other hand, the fact that FreeBSD changed and didn't suffer an
immediate backlash of breaking programs means that glibc might
consider such a change, despite the l
On Wed, Mar 22, 2023 at 03:07:01PM -0500, Eric Blake wrote:
> On Wed, Mar 22, 2023 at 04:53:42PM +0100, Laszlo Ersek wrote:
> > The solution is that glibc *too* has a bug, and that bug hides the busybox
> > bug. Namely, in glibc, going back to historical commit
t warns about the short construct (at compile- or run-time), that
would be a definitive reason to do this. But given that future
standards will require the short form to work identically to the long
form, and I'm unaware of a compiler that actually warns on the short
form, I'm 50-50 on whe
nditional,
but the condition would change to 'if (tmpdir)' instead of depending
on 'next'.
> +
> +free_tmpdir:
> + if (next == %.DEAD)
> +free (tmpdir);
And with transfer semantics and attribute cleanup, this is another
label we could elide.
> +
> +done:
perror (h->argv.ptr[0]);
> if (errno == ENOENT)
I had to check that nbd_internal_fork_safe_perror() preserves errno -
but fortunately it does (it's always a good idea for functions
designed to be used on cleanup handling paths to behave that way).
Reviewed-by: Eric Blake
--
Eric Blake
calls, which are superfluous.
>
> Signed-off-by: Laszlo Ersek
> Reviewed-by: Richard W.M. Jones
> ---
>
Reviewed-by: Eric Blake
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
_
1 - 100 of 3081 matches
Mail list logo
