[PATCH v2 00/12] Error reporting cleanup, a fix, and &error_warn removal

[PATCH v2 00/12] Error reporting cleanup, a fix, and &error_warn removal

[Markus Armbruster]

From qapi/error.h:

 * Please don't error_setg(&error_fatal, ...), use error_report() and
 * exit(), because that's more obvious.
 * Likewise, don't error_setg(&error_abort, ...), use assert().

Not mentioned, but just as undesirable: error_setg(&error_warn, ...).

This series eliminates such uses, and gets rid of &error_warn.
&error_warn has multiple issues and little use.  PATCH 12 has full
rationale.

A note on warnings: we don't use warnings much, and when we use them,
they're often pretty bad.  See my memo "Abuse of warnings for
unhandled errors and programming errors"
Message-ID: <87h5yijh3b.fsf@pond.sub.org>
https://lore.kernel.org/qemu-devel/87h5yijh3b.fsf@pond.sub.org/

v2:
* PATCH 03: Mention change of cxl_fmws_link() return value in commit
  message [Jonathan]
* PATCH 04: Change exit(1) to g_assert_not_reached(), because it's a
  programming error.
* PATCH 06+07: Replace questions in commit message by answers from
  review.
* PATCH 06: Fix a format string
* PATCH 08: Keep warnings instead of reverting to silence [Daniel]
* PATCH 12: Adjusted for replaced PATCH 08

Issues raised in review I decided not to address in this series:
* PATCH 03: messages could be improved further, in particular the
  "gdbstub: " prefix could be dropped
* ebpf_rss_load() can return false without setting an error
* Capture the discussion on how to deal with undhandled errors in
  cover letter and/or commit messages.

The first two could be done on top.

Markus Armbruster (12):
  monitor: Clean up HMP gdbserver error reporting
  tcg: Fix error reporting on mprotect() failure in  tcg_region_init()
  hw/cxl: Convert cxl_fmws_link() to Error
  migration/cpr: Clean up error reporting in cpr_resave_fd()
  hw/remote/vfio-user: Clean up error reporting
  net/slirp: Clean up error reporting
  ui/spice-core: Clean up error reporting
  util/oslib-win32: Do not treat null @errp as &error_warn
  ui/pixman: Consistent error handling in qemu_pixman_shareable_free()
  ui/dbus: Clean up dbus_update_gl_cb() error checking
  ui/dbus: Consistent handling of texture mutex failure
  error: Kill @error_warn

 include/exec/gdbstub.h         |  3 ---
 include/qapi/error.h           |  6 ------
 include/system/os-win32.h      |  5 ++++-
 hw/cxl/cxl-host.c              |  7 ++++---
 hw/display/virtio-gpu.c        |  8 ++++++--
 hw/net/virtio-net.c            |  8 +++++++-
 hw/remote/vfio-user-obj.c      |  9 +++------
 io/channel-socket.c            |  4 ++--
 io/channel-watch.c             |  6 +++---
 migration/cpr.c                |  9 +++++----
 monitor/hmp-cmds.c             |  7 ++++---
 net/slirp.c                    |  6 ++++--
 tcg/region.c                   |  8 ++++++--
 tests/unit/test-error-report.c | 17 -----------------
 ui/dbus-listener.c             | 22 +++++++++++++++-------
 ui/gtk.c                       |  6 +++++-
 ui/qemu-pixman.c               |  5 ++++-
 ui/spice-core.c                |  3 ++-
 util/aio-win32.c               |  2 +-
 util/error.c                   |  5 +----
 util/oslib-win32.c             | 26 +++++++++++++++++++++-----
 21 files changed, 97 insertions(+), 75 deletions(-)

-- 
2.49.0

[PATCH v2 01/12] monitor: Clean up HMP gdbserver error reporting

[Markus Armbruster]

HMP command gdbserver used to emit two error messages for certain
errors.  For instance, with -M none:

    (qemu) gdbserver
    gdbstub: meaningless to attach gdb to a machine without any CPU.
    Could not open gdbserver on device 'tcp::1234'

The first message is the specific error, and the second one a generic
additional message that feels superfluous to me.

Commit c0e6b8b798b (system: propagate Error to gdbserver_start (and
other device setups)) turned the first message into a warning:

    warning: gdbstub: meaningless to attach gdb to a machine without any CPU.
    Could not open gdbserver on device 'tcp::1234'

This is arguably worse.

hmp_gdbserver() passes &error_warn to gdbserver_start(), so that
failure gets reported as warning, and then additionally emits the
generic error on failure.  This is a misuse of &error_warn.

Instead, receive the error in &err and report it, as usual.  With
this, gdbserver reports just the error:

    gdbstub: meaningless to attach gdb to a machine without any CPU.

Cc: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
---
 include/exec/gdbstub.h | 3 ---
 monitor/hmp-cmds.c     | 7 ++++---
 2 files changed, 4 insertions(+), 6 deletions(-)

diff --git a/include/exec/gdbstub.h b/include/exec/gdbstub.h
index a16c0051ce..bd7182c4d3 100644
--- a/include/exec/gdbstub.h
+++ b/include/exec/gdbstub.h
@@ -55,9 +55,6 @@ void gdb_unregister_coprocessor_all(CPUState *cpu);
  * system emulation you can use a full chardev spec for your gdbserver
  * port.
  *
- * The error handle should be either &error_fatal (for start-up) or
- * &error_warn (for QMP/HMP initiated sessions).
- *
  * Returns true when server successfully started.
  */
 bool gdbserver_start(const char *port_or_device, Error **errp);
diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
index 74a0f56566..33a88ce205 100644
--- a/monitor/hmp-cmds.c
+++ b/monitor/hmp-cmds.c
@@ -280,14 +280,15 @@ void hmp_log(Monitor *mon, const QDict *qdict)
 
 void hmp_gdbserver(Monitor *mon, const QDict *qdict)
 {
+    Error *err = NULL;
     const char *device = qdict_get_try_str(qdict, "device");
+
     if (!device) {
         device = "tcp::" DEFAULT_GDBSTUB_PORT;
     }
 
-    if (!gdbserver_start(device, &error_warn)) {
-        monitor_printf(mon, "Could not open gdbserver on device '%s'\n",
-                       device);
+    if (!gdbserver_start(device, &err)) {
+        error_report_err(err);
     } else if (strcmp(device, "none") == 0) {
         monitor_printf(mon, "Disabled gdbserver\n");
     } else {
-- 
2.49.0

[PATCH v2 02/12] tcg: Fix error reporting on mprotect() failure in tcg_region_init()

[Markus Armbruster]

tcg_region_init() calls one of qemu_mprotect_rwx(),
qemu_mprotect_rw(), and mprotect(), then reports failure with
error_setg_errno(&error_fatal, errno, ...).

The use of &error_fatal is undesirable.  qapi/error.h advises:

 * Please don't error_setg(&error_fatal, ...), use error_report() and
 * exit(), because that's more obvious.

The use of errno is wrong.  qemu_mprotect_rwx() and qemu_mprotect_rw()
wrap around qemu_mprotect__osdep().  qemu_mprotect__osdep() calls
mprotect() on POSIX, VirtualProtect() on Windows, and reports failure
with error_report().  VirtualProtect() doesn't set errno.  mprotect()
does, but error_report() may clobber it.

Fix tcg_region_init() to report errors only when it calls mprotect(),
and rely on qemu_mprotect_rwx()'s and qemu_mprotect_rw()'s error
reporting otherwise.  Use error_report(), not error_setg().

Fixes: 22c6a9938f75 (tcg: Merge buffer protection and guard page protection)
Fixes: 6bc144237a85 (tcg: Use Error with alloc_code_gen_buffer)
Cc: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
---
 tcg/region.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/tcg/region.c b/tcg/region.c
index 7ea0b37a84..74e3b4b774 100644
--- a/tcg/region.c
+++ b/tcg/region.c
@@ -832,13 +832,17 @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_threads)
             } else {
 #ifdef CONFIG_POSIX
                 rc = mprotect(start, end - start, need_prot);
+                if (rc) {
+                    error_report("mprotect of jit buffer: %s",
+                                 strerror(errno));
+                }
+
 #else
                 g_assert_not_reached();
 #endif
             }
             if (rc) {
-                error_setg_errno(&error_fatal, errno,
-                                 "mprotect of jit buffer");
+                exit(1);
             }
         }
         if (have_prot != 0) {
-- 
2.49.0

Re: [PATCH v2 02/12] tcg: Fix error reporting on mprotect() failure in tcg_region_init()

[Richard Henderson]

On 9/17/25 04:51, Markus Armbruster wrote:
> tcg_region_init() calls one of qemu_mprotect_rwx(),
> qemu_mprotect_rw(), and mprotect(), then reports failure with
> error_setg_errno(&error_fatal, errno, ...).
> 
> The use of &error_fatal is undesirable.  qapi/error.h advises:
> 
>   * Please don't error_setg(&error_fatal, ...), use error_report() and
>   * exit(), because that's more obvious.
> 
> The use of errno is wrong.  qemu_mprotect_rwx() and qemu_mprotect_rw()
> wrap around qemu_mprotect__osdep().  qemu_mprotect__osdep() calls
> mprotect() on POSIX, VirtualProtect() on Windows, and reports failure
> with error_report().  VirtualProtect() doesn't set errno.  mprotect()
> does, but error_report() may clobber it.
> 
> Fix tcg_region_init() to report errors only when it calls mprotect(),
> and rely on qemu_mprotect_rwx()'s and qemu_mprotect_rw()'s error
> reporting otherwise.  Use error_report(), not error_setg().
> 
> Fixes: 22c6a9938f75 (tcg: Merge buffer protection and guard page protection)
> Fixes: 6bc144237a85 (tcg: Use Error with alloc_code_gen_buffer)
> Cc: Richard Henderson <richard.henderson@linaro.org>
> Signed-off-by: Markus Armbruster <armbru@redhat.com>
> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
> ---
>   tcg/region.c | 8 ++++++--
>   1 file changed, 6 insertions(+), 2 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: [PATCH v2 02/12] tcg: Fix error reporting on mprotect() failure in tcg_region_init()

[Akihiko Odaki]

On 2025/09/17 20:51, Markus Armbruster wrote:
> tcg_region_init() calls one of qemu_mprotect_rwx(),
> qemu_mprotect_rw(), and mprotect(), then reports failure with
> error_setg_errno(&error_fatal, errno, ...).
> 
> The use of &error_fatal is undesirable.  qapi/error.h advises:
> 
>   * Please don't error_setg(&error_fatal, ...), use error_report() and
>   * exit(), because that's more obvious.
> 
> The use of errno is wrong.  qemu_mprotect_rwx() and qemu_mprotect_rw()
> wrap around qemu_mprotect__osdep().  qemu_mprotect__osdep() calls
> mprotect() on POSIX, VirtualProtect() on Windows, and reports failure
> with error_report().  VirtualProtect() doesn't set errno.  mprotect()
> does, but error_report() may clobber it.
> 
> Fix tcg_region_init() to report errors only when it calls mprotect(),
> and rely on qemu_mprotect_rwx()'s and qemu_mprotect_rw()'s error
> reporting otherwise.  Use error_report(), not error_setg().
> 
> Fixes: 22c6a9938f75 (tcg: Merge buffer protection and guard page protection)
> Fixes: 6bc144237a85 (tcg: Use Error with alloc_code_gen_buffer)
> Cc: Richard Henderson <richard.henderson@linaro.org>
> Signed-off-by: Markus Armbruster <armbru@redhat.com>
> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
> ---
>   tcg/region.c | 8 ++++++--
>   1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/tcg/region.c b/tcg/region.c
> index 7ea0b37a84..74e3b4b774 100644
> --- a/tcg/region.c
> +++ b/tcg/region.c
> @@ -832,13 +832,17 @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_threads)
>               } else {
>   #ifdef CONFIG_POSIX
>                   rc = mprotect(start, end - start, need_prot);
> +                if (rc) {
> +                    error_report("mprotect of jit buffer: %s",
> +                                 strerror(errno));
> +                }
> +

Nitpick: this blank line at the end of #ifdef looks strange to me.

>   #else
>                   g_assert_not_reached();
>   #endif
>               }
>               if (rc) {
> -                error_setg_errno(&error_fatal, errno,
> -                                 "mprotect of jit buffer");
> +                exit(1);
>               }
>           }
>           if (have_prot != 0) {

Re: [PATCH v2 02/12] tcg: Fix error reporting on mprotect() failure in tcg_region_init()

[Markus Armbruster]

Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp> writes:

> On 2025/09/17 20:51, Markus Armbruster wrote:
>> tcg_region_init() calls one of qemu_mprotect_rwx(),
>> qemu_mprotect_rw(), and mprotect(), then reports failure with
>> error_setg_errno(&error_fatal, errno, ...).
>> The use of &error_fatal is undesirable.  qapi/error.h advises:
>>   * Please don't error_setg(&error_fatal, ...), use error_report() and
>>   * exit(), because that's more obvious.
>> The use of errno is wrong.  qemu_mprotect_rwx() and qemu_mprotect_rw()
>> wrap around qemu_mprotect__osdep().  qemu_mprotect__osdep() calls
>> mprotect() on POSIX, VirtualProtect() on Windows, and reports failure
>> with error_report().  VirtualProtect() doesn't set errno.  mprotect()
>> does, but error_report() may clobber it.
>> Fix tcg_region_init() to report errors only when it calls mprotect(),
>> and rely on qemu_mprotect_rwx()'s and qemu_mprotect_rw()'s error
>> reporting otherwise.  Use error_report(), not error_setg().
>> Fixes: 22c6a9938f75 (tcg: Merge buffer protection and guard page protection)
>> Fixes: 6bc144237a85 (tcg: Use Error with alloc_code_gen_buffer)
>> Cc: Richard Henderson <richard.henderson@linaro.org>
>> Signed-off-by: Markus Armbruster <armbru@redhat.com>
>> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
>> ---
>>   tcg/region.c | 8 ++++++--
>>   1 file changed, 6 insertions(+), 2 deletions(-)
>> diff --git a/tcg/region.c b/tcg/region.c
>> index 7ea0b37a84..74e3b4b774 100644
>> --- a/tcg/region.c
>> +++ b/tcg/region.c
>> @@ -832,13 +832,17 @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_threads)
>>               } else {
>>   #ifdef CONFIG_POSIX
>>                   rc = mprotect(start, end - start, need_prot);
>> +                if (rc) {
>> +                    error_report("mprotect of jit buffer: %s",
>> +                                 strerror(errno));
>> +                }
>> +
>
> Nitpick: this blank line at the end of #ifdef looks strange to me.

Accident, will drop it.  Thanks!

>>   #else
>>                   g_assert_not_reached();
>>   #endif
>>               }
>>               if (rc) {
>> -                error_setg_errno(&error_fatal, errno,
>> -                                 "mprotect of jit buffer");
>> +                exit(1);
>>               }
>>           }
>>           if (have_prot != 0) {

[PATCH v2 03/12] hw/cxl: Convert cxl_fmws_link() to Error

[Markus Armbruster]

Functions that use an Error **errp parameter to return errors should
not also report them to the user, because reporting is the caller's
job.  When the caller does, the error is reported twice.  When it
doesn't (because it recovered from the error), there is no error to
report, i.e. the report is bogus.

cxl_fmws_link_targets() violates this principle: it calls
error_setg(&error_fatal, ...) via cxl_fmws_link().  Goes back to
commit 584f722eb3ab (hw/cxl: Make the CXL fixed memory windows
devices.)  Currently harmless, because cxl_fmws_link_targets()'s
callers always pass &error_fatal.  Clean this up by converting
cxl_fmws_link() to Error.

Also change its return value on error from 1 to -1 to conform to the
rules laid in qapi/error.h.  It's call chain cxl_fmws_link_targets()
via object_child_foreach_recursive() is fine with that.

Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
---
 hw/cxl/cxl-host.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/hw/cxl/cxl-host.c b/hw/cxl/cxl-host.c
index 5c2ce25a19..0d891c651d 100644
--- a/hw/cxl/cxl-host.c
+++ b/hw/cxl/cxl-host.c
@@ -72,6 +72,7 @@ static void cxl_fixed_memory_window_config(CXLFixedMemoryWindowOptions *object,
 
 static int cxl_fmws_link(Object *obj, void *opaque)
 {
+    Error **errp = opaque;
     struct CXLFixedWindow *fw;
     int i;
 
@@ -87,9 +88,9 @@ static int cxl_fmws_link(Object *obj, void *opaque)
         o = object_resolve_path_type(fw->targets[i], TYPE_PXB_CXL_DEV,
                                      &ambig);
         if (!o) {
-            error_setg(&error_fatal, "Could not resolve CXLFM target %s",
+            error_setg(errp, "Could not resolve CXLFM target %s",
                        fw->targets[i]);
-            return 1;
+            return -1;
         }
         fw->target_hbs[i] = PXB_CXL_DEV(o);
     }
@@ -99,7 +100,7 @@ static int cxl_fmws_link(Object *obj, void *opaque)
 void cxl_fmws_link_targets(Error **errp)
 {
     /* Order doesn't matter for this, so no need to build list */
-    object_child_foreach_recursive(object_get_root(), cxl_fmws_link, NULL);
+    object_child_foreach_recursive(object_get_root(), cxl_fmws_link, errp);
 }
 
 static bool cxl_hdm_find_target(uint32_t *cache_mem, hwaddr addr,
-- 
2.49.0

Re: [PATCH v2 03/12] hw/cxl: Convert cxl_fmws_link() to Error

[Jonathan Cameron via]

On Wed, 17 Sep 2025 13:51:58 +0200
Markus Armbruster <armbru@redhat.com> wrote:

> Functions that use an Error **errp parameter to return errors should
> not also report them to the user, because reporting is the caller's
> job.  When the caller does, the error is reported twice.  When it
> doesn't (because it recovered from the error), there is no error to
> report, i.e. the report is bogus.
> 
> cxl_fmws_link_targets() violates this principle: it calls
> error_setg(&error_fatal, ...) via cxl_fmws_link().  Goes back to
> commit 584f722eb3ab (hw/cxl: Make the CXL fixed memory windows
> devices.)  Currently harmless, because cxl_fmws_link_targets()'s
> callers always pass &error_fatal.  Clean this up by converting
> cxl_fmws_link() to Error.
> 
> Also change its return value on error from 1 to -1 to conform to the
> rules laid in qapi/error.h.  It's call chain cxl_fmws_link_targets()
> via object_child_foreach_recursive() is fine with that.
> 
> Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
> Signed-off-by: Markus Armbruster <armbru@redhat.com>
> Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
LGTM with the updated commit message.  Thanks!

Jonathan

[PATCH v2 04/12] migration/cpr: Clean up error reporting in cpr_resave_fd()

[Markus Armbruster]

qapi/error.h advises:

 * Please don't error_setg(&error_fatal, ...), use error_report() and
 * exit(), because that's more obvious.

Do that, and replace exit() by g_assert_not_reached(), since this is
actually a programming error.

Cc: Steve Sistare <steven.sistare@oracle.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Steve Sistare <steven.sistare@oracle.com>
---
 migration/cpr.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/migration/cpr.c b/migration/cpr.c
index 42ad0b0d50..9848a21ea6 100644
--- a/migration/cpr.c
+++ b/migration/cpr.c
@@ -7,6 +7,7 @@
 
 #include "qemu/osdep.h"
 #include "qapi/error.h"
+#include "qemu/error-report.h"
 #include "hw/vfio/vfio-device.h"
 #include "migration/cpr.h"
 #include "migration/misc.h"
@@ -100,10 +101,10 @@ void cpr_resave_fd(const char *name, int id, int fd)
     if (old_fd < 0) {
         cpr_save_fd(name, id, fd);
     } else if (old_fd != fd) {
-        error_setg(&error_fatal,
-                   "internal error: cpr fd '%s' id %d value %d "
-                   "already saved with a different value %d",
-                   name, id, fd, old_fd);
+        error_report("internal error: cpr fd '%s' id %d value %d "
+                     "already saved with a different value %d",
+                     name, id, fd, old_fd);
+        g_assert_not_reached();
     }
 }
 
-- 
2.49.0

[PATCH v2 05/12] hw/remote/vfio-user: Clean up error reporting

[Markus Armbruster]

VFU_OBJECT_ERROR() reports the error with error_setg(&error_abort,
...) when auto-shutdown is enabled, else with error_report().

Issues:

1. The error is serious enough to warrant aborting the process when
auto-shutdown is enabled, yet harmless enough to permit carrying on
when it's disabled.  This makes no sense to me.

2. Like assert(), &error_abort is strictly for programming errors.  Is
this one?  Or should we exit(1) instead?

3. qapi/error.h advises "don't error_setg(&error_abort, ...), use
assert()."

This patch addresses just 3.

Cc: Jagannathan Raman <jag.raman@oracle.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 hw/remote/vfio-user-obj.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/hw/remote/vfio-user-obj.c b/hw/remote/vfio-user-obj.c
index ea6165ebdc..eb96982a3a 100644
--- a/hw/remote/vfio-user-obj.c
+++ b/hw/remote/vfio-user-obj.c
@@ -75,12 +75,9 @@ OBJECT_DECLARE_TYPE(VfuObject, VfuObjectClass, VFU_OBJECT)
  */
 #define VFU_OBJECT_ERROR(o, fmt, ...)                                     \
     {                                                                     \
-        if (vfu_object_auto_shutdown()) {                                 \
-            error_setg(&error_abort, (fmt), ## __VA_ARGS__);              \
-        } else {                                                          \
-            error_report((fmt), ## __VA_ARGS__);                          \
-        }                                                                 \
-    }                                                                     \
+        error_report((fmt), ## __VA_ARGS__);                              \
+        assert(!vfu_object_auto_shutdown());                              \
+    }
 
 struct VfuObjectClass {
     ObjectClass parent_class;
-- 
2.49.0

[PATCH v2 06/12] net/slirp: Clean up error reporting

[Markus Armbruster]

net_slirp_register_poll_sock() and net_slirp_unregister_poll_sock()
report WSAEventSelect() failure with error_setg(&error_warn, ...).
error_setg_win32(&error_warn, ...) is undesirable just like
error_setg(&error_fatal, ...) and error_setg(&error_abort, ...)  are.
Replace by warn_report().

The failures should probably be errors, but these functions implement
callbacks that cannot fail, exit(1) would be too harsh, and silent
failure we don't want.  Thus, warnings.

Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 net/slirp.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/net/slirp.c b/net/slirp.c
index 9657e86a84..62f2684609 100644
--- a/net/slirp.c
+++ b/net/slirp.c
@@ -262,7 +262,8 @@ static void net_slirp_register_poll_sock(slirp_os_socket fd, void *opaque)
     if (WSAEventSelect(fd, event_notifier_get_handle(&ctxt->notifier),
                        FD_READ | FD_ACCEPT | FD_CLOSE |
                        FD_CONNECT | FD_WRITE | FD_OOB) != 0) {
-        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
+        warn_report("failed to WSAEventSelect(): %s",
+                    g_win32_error_message(WSAGetLastError()));
     }
 #endif
 }
@@ -271,7 +272,8 @@ static void net_slirp_unregister_poll_sock(slirp_os_socket fd, void *opaque)
 {
 #ifdef WIN32
     if (WSAEventSelect(fd, NULL, 0) != 0) {
-        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
+        warn_report("failed to WSAEventSelect(): %s",
+                    g_win32_error_message(WSAGetLastError()));
     }
 #endif
 }
-- 
2.49.0

Re: [PATCH v2 06/12] net/slirp: Clean up error reporting

[Akihiko Odaki]

On 2025/09/17 20:52, Markus Armbruster wrote:
> net_slirp_register_poll_sock() and net_slirp_unregister_poll_sock()
> report WSAEventSelect() failure with error_setg(&error_warn, ...).
> error_setg_win32(&error_warn, ...) is undesirable just like
> error_setg(&error_fatal, ...) and error_setg(&error_abort, ...)  are.
> Replace by warn_report().
> 
> The failures should probably be errors, but these functions implement
> callbacks that cannot fail, exit(1) would be too harsh, and silent
> failure we don't want.  Thus, warnings.
> 
> Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
> Signed-off-by: Markus Armbruster <armbru@redhat.com>
> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> ---
>   net/slirp.c | 6 ++++--
>   1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/net/slirp.c b/net/slirp.c
> index 9657e86a84..62f2684609 100644
> --- a/net/slirp.c
> +++ b/net/slirp.c
> @@ -262,7 +262,8 @@ static void net_slirp_register_poll_sock(slirp_os_socket fd, void *opaque)
>       if (WSAEventSelect(fd, event_notifier_get_handle(&ctxt->notifier),
>                          FD_READ | FD_ACCEPT | FD_CLOSE |
>                          FD_CONNECT | FD_WRITE | FD_OOB) != 0) {
> -        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
> +        warn_report("failed to WSAEventSelect(): %s",
> +                    g_win32_error_message(WSAGetLastError()));

https://docs.gtk.org/glib-win32/func.error_message.html says:
 > The caller of the function takes ownership of the data, and is
 > responsible for freeing it.

>       }
>   #endif
>   }
> @@ -271,7 +272,8 @@ static void net_slirp_unregister_poll_sock(slirp_os_socket fd, void *opaque)
>   {
>   #ifdef WIN32
>       if (WSAEventSelect(fd, NULL, 0) != 0) {
> -        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
> +        warn_report("failed to WSAEventSelect(): %s",
> +                    g_win32_error_message(WSAGetLastError()));
>       }
>   #endif
>   }

Re: [PATCH v2 06/12] net/slirp: Clean up error reporting

[Markus Armbruster]

Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp> writes:

> On 2025/09/17 20:52, Markus Armbruster wrote:
>> net_slirp_register_poll_sock() and net_slirp_unregister_poll_sock()
>> report WSAEventSelect() failure with error_setg(&error_warn, ...).
>> error_setg_win32(&error_warn, ...) is undesirable just like
>> error_setg(&error_fatal, ...) and error_setg(&error_abort, ...)  are.
>> Replace by warn_report().
>> The failures should probably be errors, but these functions implement
>> callbacks that cannot fail, exit(1) would be too harsh, and silent
>> failure we don't want.  Thus, warnings.
>> Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
>> Signed-off-by: Markus Armbruster <armbru@redhat.com>
>> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
>> ---
>>   net/slirp.c | 6 ++++--
>>   1 file changed, 4 insertions(+), 2 deletions(-)
>> diff --git a/net/slirp.c b/net/slirp.c
>> index 9657e86a84..62f2684609 100644
>> --- a/net/slirp.c
>> +++ b/net/slirp.c
>> @@ -262,7 +262,8 @@ static void net_slirp_register_poll_sock(slirp_os_socket fd, void *opaque)
>>       if (WSAEventSelect(fd, event_notifier_get_handle(&ctxt->notifier),
>>                          FD_READ | FD_ACCEPT | FD_CLOSE |
>>                          FD_CONNECT | FD_WRITE | FD_OOB) != 0) {
>> -        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
>> +        warn_report("failed to WSAEventSelect(): %s",
>> +                    g_win32_error_message(WSAGetLastError()));
>
> https://docs.gtk.org/glib-win32/func.error_message.html says:
>> The caller of the function takes ownership of the data, and is
>> responsible for freeing it.

I'll fix it.  Thanks!

>
>>       }
>>   #endif
>>   }
>> @@ -271,7 +272,8 @@ static void net_slirp_unregister_poll_sock(slirp_os_socket fd, void *opaque)
>>   {
>>   #ifdef WIN32
>>       if (WSAEventSelect(fd, NULL, 0) != 0) {
>> -        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
>> +        warn_report("failed to WSAEventSelect(): %s",
>> +                    g_win32_error_message(WSAGetLastError()));
>>       }
>>   #endif
>>   }

[PATCH v2 07/12] ui/spice-core: Clean up error reporting

[Markus Armbruster]

watch_add() reports _open_osfhandle() failure with
error_setg(&error_warn, ...).  error_setg_win32(&error_warn, ...) is
undesirable just like error_setg(&error_fatal, ...) and
error_setg(&error_abort, ...) are.  Replace by warn_report().

The failure should probably be an error, but this function implements
a callback that doesn't take Error **.  I believe the failure will
make spice_server_init() fail in qemu_spice_init(), which is treated
as a fatal error.  The warning here provides more detail than the
error message there.

Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 ui/spice-core.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/ui/spice-core.c b/ui/spice-core.c
index 5992f9daec..97bdd171cd 100644
--- a/ui/spice-core.c
+++ b/ui/spice-core.c
@@ -132,7 +132,8 @@ static SpiceWatch *watch_add(int fd, int event_mask, SpiceWatchFunc func, void *
 #ifdef WIN32
     fd = _open_osfhandle(fd, _O_BINARY);
     if (fd < 0) {
-        error_setg_win32(&error_warn, WSAGetLastError(), "Couldn't associate a FD with the SOCKET");
+        warn_report("Couldn't associate a FD with the SOCKET: %s"
+                    g_win32_error_message(WSAGetLastError()));
         return NULL;
     }
 #endif
-- 
2.49.0

Re: [PATCH v2 07/12] ui/spice-core: Clean up error reporting

[Richard Henderson]

On 9/17/25 04:52, Markus Armbruster wrote:
> watch_add() reports _open_osfhandle() failure with
> error_setg(&error_warn, ...).  error_setg_win32(&error_warn, ...) is
> undesirable just like error_setg(&error_fatal, ...) and
> error_setg(&error_abort, ...) are.  Replace by warn_report().
> 
> The failure should probably be an error, but this function implements
> a callback that doesn't take Error **.  I believe the failure will
> make spice_server_init() fail in qemu_spice_init(), which is treated
> as a fatal error.  The warning here provides more detail than the
> error message there.
> 
> Cc: Marc-André Lureau<marcandre.lureau@redhat.com>
> Signed-off-by: Markus Armbruster<armbru@redhat.com>
> Reviewed-by: Marc-André Lureau<marcandre.lureau@redhat.com>
> ---
>   ui/spice-core.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v2 08/12] util/oslib-win32: Do not treat null @errp as &error_warn

[Markus Armbruster]

qemu_socket_select() and its wrapper qemu_socket_unselect() treat a
null @errp as &error_warn.  This is wildly inappropriate.  A caller
passing null @errp specifies that errors are to be ignored.  If
warnings are wanted, the caller must pass &error_warn.

Change callers to do that, and drop the inappropriate treatment of
null @errp.

This assumes that warnings are wanted.  I'm not familiar with the
calling code, so I can't say whether it will work when the socket is
invalid, or WSAEventSelect() fails.  If it doesn't, then this should
be an error instead of a warning.  Invalid socket might even be a
programming error.

These warnings were introduced in commit f5fd677ae7cf (win32/socket:
introduce qemu_socket_select() helper).  I considered reverting to
silence, but Daniel Berrangé asked for the warnings to be preserved.

Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 io/channel-socket.c | 4 ++--
 io/channel-watch.c  | 2 +-
 util/aio-win32.c    | 2 +-
 util/oslib-win32.c  | 7 ++-----
 4 files changed, 6 insertions(+), 9 deletions(-)

diff --git a/io/channel-socket.c b/io/channel-socket.c
index 3b7ca924ff..56ef353cd3 100644
--- a/io/channel-socket.c
+++ b/io/channel-socket.c
@@ -454,7 +454,7 @@ static void qio_channel_socket_finalize(Object *obj)
             }
         }
 #ifdef WIN32
-        qemu_socket_unselect(ioc->fd, NULL);
+        qemu_socket_unselect(ioc->fd, &error_warn);
 #endif
         close(ioc->fd);
         ioc->fd = -1;
@@ -890,7 +890,7 @@ qio_channel_socket_close(QIOChannel *ioc,
 
     if (sioc->fd != -1) {
 #ifdef WIN32
-        qemu_socket_unselect(sioc->fd, NULL);
+        qemu_socket_unselect(sioc->fd, &error_warn);
 #endif
         if (qio_channel_has_feature(ioc, QIO_CHANNEL_FEATURE_LISTEN)) {
             socket_listen_cleanup(sioc->fd, errp);
diff --git a/io/channel-watch.c b/io/channel-watch.c
index 64b486e378..ec76bd1ec6 100644
--- a/io/channel-watch.c
+++ b/io/channel-watch.c
@@ -283,7 +283,7 @@ GSource *qio_channel_create_socket_watch(QIOChannel *ioc,
 
     qemu_socket_select(sockfd, ioc->event,
                        FD_READ | FD_ACCEPT | FD_CLOSE |
-                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
+                       FD_CONNECT | FD_WRITE | FD_OOB, &error_warn);
 
     source = g_source_new(&qio_channel_socket_source_funcs,
                           sizeof(QIOChannelSocketSource));
diff --git a/util/aio-win32.c b/util/aio-win32.c
index 6583d5c5f3..b125924433 100644
--- a/util/aio-win32.c
+++ b/util/aio-win32.c
@@ -121,7 +121,7 @@ void aio_set_fd_handler(AioContext *ctx,
 
         QLIST_INSERT_HEAD_RCU(&ctx->aio_handlers, node, node);
         event = event_notifier_get_handle(&ctx->notifier);
-        qemu_socket_select(fd, event, bitmask, NULL);
+        qemu_socket_select(fd, event, bitmask, &error_warn);
     }
     if (old_node) {
         aio_remove_fd_handler(ctx, old_node);
diff --git a/util/oslib-win32.c b/util/oslib-win32.c
index b7351634ec..df118b164e 100644
--- a/util/oslib-win32.c
+++ b/util/oslib-win32.c
@@ -180,7 +180,8 @@ static int socket_error(void)
 void qemu_socket_set_block(int fd)
 {
     unsigned long opt = 0;
-    qemu_socket_unselect(fd, NULL);
+
+    qemu_socket_unselect(fd, &error_warn);
     ioctlsocket(fd, FIONBIO, &opt);
 }
 
@@ -296,10 +297,6 @@ bool qemu_socket_select(int sockfd, WSAEVENT hEventObject,
 {
     SOCKET s = _get_osfhandle(sockfd);
 
-    if (errp == NULL) {
-        errp = &error_warn;
-    }
-
     if (s == INVALID_SOCKET) {
         error_setg(errp, "invalid socket fd=%d", sockfd);
         return false;
-- 
2.49.0

Re: [PATCH v2 08/12] util/oslib-win32: Do not treat null @errp as &error_warn

[Richard Henderson]

On 9/17/25 04:52, Markus Armbruster wrote:
> qemu_socket_select() and its wrapper qemu_socket_unselect() treat a
> null @errp as &error_warn.  This is wildly inappropriate.  A caller
> passing null @errp specifies that errors are to be ignored.  If
> warnings are wanted, the caller must pass &error_warn.
> 
> Change callers to do that, and drop the inappropriate treatment of
> null @errp.
> 
> This assumes that warnings are wanted.  I'm not familiar with the
> calling code, so I can't say whether it will work when the socket is
> invalid, or WSAEventSelect() fails.  If it doesn't, then this should
> be an error instead of a warning.  Invalid socket might even be a
> programming error.
> 
> These warnings were introduced in commit f5fd677ae7cf (win32/socket:
> introduce qemu_socket_select() helper).  I considered reverting to
> silence, but Daniel Berrangé asked for the warnings to be preserved.
> 
> Cc: Marc-André Lureau<marcandre.lureau@redhat.com>
> Signed-off-by: Markus Armbruster<armbru@redhat.com>
> ---
>   io/channel-socket.c | 4 ++--
>   io/channel-watch.c  | 2 +-
>   util/aio-win32.c    | 2 +-
>   util/oslib-win32.c  | 7 ++-----
>   4 files changed, 6 insertions(+), 9 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[PATCH v2 09/12] ui/pixman: Consistent error handling in qemu_pixman_shareable_free()

[Markus Armbruster]

qemu_pixman_shareable_free() wraps around either qemu_memfd_free() or
qemu_win32_map_free().  The former reports trouble as error, with
error_report(), then succeeds.  The latter reports it as warning (we
pass it &error_warn), then succeeds.

Change the latter to report as error, too.

Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 ui/qemu-pixman.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/ui/qemu-pixman.c b/ui/qemu-pixman.c
index ef4e71da11..e46c6232cf 100644
--- a/ui/qemu-pixman.c
+++ b/ui/qemu-pixman.c
@@ -288,7 +288,10 @@ qemu_pixman_shareable_free(qemu_pixman_shareable handle,
                            void *ptr, size_t size)
 {
 #ifdef WIN32
-    qemu_win32_map_free(ptr, handle, &error_warn);
+    Error *err = NULL;
+
+    qemu_win32_map_free(ptr, handle, &err);
+    error_report_err(err);
 #else
     qemu_memfd_free(ptr, size, handle);
 #endif
-- 
2.49.0

[PATCH v2 10/12] ui/dbus: Clean up dbus_update_gl_cb() error checking

[Markus Armbruster]

From GLib "Rules for use of GError":

    A GError* must be initialized to NULL before passing its address
    to a function that can report errors.

dbus_update_gl_cb() seemingly violates this rule: it passes &err to
qemu_dbus_display1_listener_call_update_dmabuf_finish() and to
qemu_dbus_display1_listener_win32_d3d11_call_update_texture2d_finish()
without clearing it in between.  Harmless, because the first call is
guarded by #ifdef CONFIG_GBM, the second by #ifdef WIN32, and the two
are mutually exclusive.  I think.

Clean this up to be obviously correct.

Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 ui/dbus-listener.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/ui/dbus-listener.c b/ui/dbus-listener.c
index 42875b8eed..09d7a319b1 100644
--- a/ui/dbus-listener.c
+++ b/ui/dbus-listener.c
@@ -221,18 +221,21 @@ static void dbus_update_gl_cb(GObject *source_object,
 #ifdef CONFIG_GBM
     success = qemu_dbus_display1_listener_call_update_dmabuf_finish(
         ddl->proxy, res, &err);
+    if (!success) {
+        error_report("Failed to call update: %s", err->message);
+    }
 #endif
 
 #ifdef WIN32
     success = qemu_dbus_display1_listener_win32_d3d11_call_update_texture2d_finish(
         ddl->d3d11_proxy, res, &err);
-    d3d_texture2d_acquire0(ddl->d3d_texture, &error_warn);
-#endif
-
     if (!success) {
         error_report("Failed to call update: %s", err->message);
     }
 
+    d3d_texture2d_acquire0(ddl->d3d_texture, &error_warn);
+#endif
+
     graphic_hw_gl_block(ddl->dcl.con, false);
     g_object_unref(ddl);
 }
-- 
2.49.0

[PATCH v2 11/12] ui/dbus: Consistent handling of texture mutex failure

[Markus Armbruster]

We report d3d_texture2d_acquire0() and d3d_texture2d_release0()
failure as error, except in dbus_update_gl_cb(), where we report it as
warning.  Report it as error there as well.

Cc: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 ui/dbus-listener.c | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/ui/dbus-listener.c b/ui/dbus-listener.c
index 09d7a319b1..b82e7c7115 100644
--- a/ui/dbus-listener.c
+++ b/ui/dbus-listener.c
@@ -214,26 +214,31 @@ static void dbus_update_gl_cb(GObject *source_object,
                               GAsyncResult *res,
                               gpointer user_data)
 {
-    g_autoptr(GError) err = NULL;
+    g_autoptr(GError) gerr = NULL;
+#ifdef WIN32
+    Error **err = NULL;
+#endif
     DBusDisplayListener *ddl = user_data;
     bool success;
 
 #ifdef CONFIG_GBM
     success = qemu_dbus_display1_listener_call_update_dmabuf_finish(
-        ddl->proxy, res, &err);
+        ddl->proxy, res, &gerr);
     if (!success) {
-        error_report("Failed to call update: %s", err->message);
+        error_report("Failed to call update: %s", gerr->message);
     }
 #endif
 
 #ifdef WIN32
     success = qemu_dbus_display1_listener_win32_d3d11_call_update_texture2d_finish(
-        ddl->d3d11_proxy, res, &err);
+        ddl->d3d11_proxy, res, &gerr);
     if (!success) {
-        error_report("Failed to call update: %s", err->message);
+        error_report("Failed to call update: %s", gerr->message);
     }
 
-    d3d_texture2d_acquire0(ddl->d3d_texture, &error_warn);
+    if (!d3d_texture2d_acquire0(ddl->d3d_texture, &err)) {
+        error_report_err(err);
+    }
 #endif
 
     graphic_hw_gl_block(ddl->dcl.con, false);
-- 
2.49.0

[PATCH v2 12/12] error: Kill @error_warn

[Markus Armbruster]

We added @error_warn some two years ago in commit 3ffef1a55ca (error:
add global &error_warn destination).  It has multiple issues:

* error.h's big comment was not updated for it.

* Function contracts were not updated for it.

* ERRP_GUARD() is unaware of @error_warn, and fails to mask it from
  error_prepend() and such.  These crash on @error_warn, as pointed
  out by Akihiko Odaki.

All fixable.  However, after more than two years, we had just of 15
uses, of which the last few patches removed seven as unclean or
otherwise undesirable, adding back five elsewhere.  I didn't look
closely enough at the remaining seven to decide whether they are
desirable or not.

I don't think this feature earns its keep.  Drop it.

Thanks-to: Akihiko  Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 include/qapi/error.h           |  6 ------
 include/system/os-win32.h      |  5 ++++-
 hw/display/virtio-gpu.c        |  8 ++++++--
 hw/net/virtio-net.c            |  8 +++++++-
 io/channel-socket.c            |  4 ++--
 io/channel-watch.c             |  6 +++---
 tests/unit/test-error-report.c | 17 -----------------
 ui/gtk.c                       |  6 +++++-
 util/aio-win32.c               |  2 +-
 util/error.c                   |  5 +----
 util/oslib-win32.c             | 21 ++++++++++++++++++++-
 11 files changed, 49 insertions(+), 39 deletions(-)

diff --git a/include/qapi/error.h b/include/qapi/error.h
index 41e3816380..b16c6303f8 100644
--- a/include/qapi/error.h
+++ b/include/qapi/error.h
@@ -533,12 +533,6 @@ static inline void error_propagator_cleanup(ErrorPropagator *prop)
 
 G_DEFINE_AUTO_CLEANUP_CLEAR_FUNC(ErrorPropagator, error_propagator_cleanup);
 
-/*
- * Special error destination to warn on error.
- * See error_setg() and error_propagate() for details.
- */
-extern Error *error_warn;
-
 /*
  * Special error destination to abort on error.
  * See error_setg() and error_propagate() for details.
diff --git a/include/system/os-win32.h b/include/system/os-win32.h
index 3aa6cee4c2..22d72babdf 100644
--- a/include/system/os-win32.h
+++ b/include/system/os-win32.h
@@ -168,11 +168,14 @@ static inline void qemu_funlockfile(FILE *f)
 #endif
 }
 
-/* Helper for WSAEventSelect, to report errors */
+/* Helpers for WSAEventSelect() */
 bool qemu_socket_select(int sockfd, WSAEVENT hEventObject,
                         long lNetworkEvents, Error **errp);
+void qemu_socket_select_nofail(int sockfd, WSAEVENT hEventObject,
+                               long lNetworkEvents);
 
 bool qemu_socket_unselect(int sockfd, Error **errp);
+void qemu_socket_unselect_nofail(int sockfd);
 
 /* We wrap all the sockets functions so that we can set errno based on
  * WSAGetLastError(), and use file-descriptors instead of SOCKET.
diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
index 0a1a625b0e..de35902213 100644
--- a/hw/display/virtio-gpu.c
+++ b/hw/display/virtio-gpu.c
@@ -242,6 +242,7 @@ static uint32_t calc_image_hostmem(pixman_format_code_t pformat,
 static void virtio_gpu_resource_create_2d(VirtIOGPU *g,
                                           struct virtio_gpu_ctrl_command *cmd)
 {
+    Error *err = NULL;
     pixman_format_code_t pformat;
     struct virtio_gpu_simple_resource *res;
     struct virtio_gpu_resource_create_2d c2d;
@@ -293,7 +294,8 @@ static void virtio_gpu_resource_create_2d(VirtIOGPU *g,
                 c2d.width,
                 c2d.height,
                 c2d.height ? res->hostmem / c2d.height : 0,
-                &error_warn)) {
+                &err)) {
+            warn_report_err(err);
             goto end;
         }
     }
@@ -1282,6 +1284,7 @@ static int virtio_gpu_load(QEMUFile *f, void *opaque, size_t size,
                            const VMStateField *field)
 {
     VirtIOGPU *g = opaque;
+    Error *err = NULL;
     struct virtio_gpu_simple_resource *res;
     uint32_t resource_id, pformat;
     int i;
@@ -1317,7 +1320,8 @@ static int virtio_gpu_load(QEMUFile *f, void *opaque, size_t size,
                                              res->width,
                                              res->height,
                                              res->height ? res->hostmem / res->height : 0,
-                                             &error_warn)) {
+                                             &err)) {
+            warn_report_err(err);
             g_free(res);
             return -EINVAL;
         }
diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
index 6b5b5dace3..7848e26278 100644
--- a/hw/net/virtio-net.c
+++ b/hw/net/virtio-net.c
@@ -1289,6 +1289,8 @@ exit:
 
 static bool virtio_net_load_ebpf(VirtIONet *n, Error **errp)
 {
+    Error *err = NULL;
+
     if (!virtio_net_attach_ebpf_to_backend(n->nic, -1)) {
         return true;
     }
@@ -1306,7 +1308,11 @@ static bool virtio_net_load_ebpf(VirtIONet *n, Error **errp)
         return virtio_net_load_ebpf_fds(n, errp);
     }
 
-    ebpf_rss_load(&n->ebpf_rss, &error_warn);
+    ebpf_rss_load(&n->ebpf_rss, &err);
+    /* Beware, ebpf_rss_load() can return false with @err unset */
+    if (err) {
+        warn_report_err(err);
+    }
     return true;
 }
 
diff --git a/io/channel-socket.c b/io/channel-socket.c
index 56ef353cd3..c21592d141 100644
--- a/io/channel-socket.c
+++ b/io/channel-socket.c
@@ -454,7 +454,7 @@ static void qio_channel_socket_finalize(Object *obj)
             }
         }
 #ifdef WIN32
-        qemu_socket_unselect(ioc->fd, &error_warn);
+        qemu_socket_unselect_nofail(ioc->fd);
 #endif
         close(ioc->fd);
         ioc->fd = -1;
@@ -890,7 +890,7 @@ qio_channel_socket_close(QIOChannel *ioc,
 
     if (sioc->fd != -1) {
 #ifdef WIN32
-        qemu_socket_unselect(sioc->fd, &error_warn);
+        qemu_socket_unselect_nofail(sioc->fd);
 #endif
         if (qio_channel_has_feature(ioc, QIO_CHANNEL_FEATURE_LISTEN)) {
             socket_listen_cleanup(sioc->fd, errp);
diff --git a/io/channel-watch.c b/io/channel-watch.c
index ec76bd1ec6..018648b36b 100644
--- a/io/channel-watch.c
+++ b/io/channel-watch.c
@@ -281,9 +281,9 @@ GSource *qio_channel_create_socket_watch(QIOChannel *ioc,
     GSource *source;
     QIOChannelSocketSource *ssource;
 
-    qemu_socket_select(sockfd, ioc->event,
-                       FD_READ | FD_ACCEPT | FD_CLOSE |
-                       FD_CONNECT | FD_WRITE | FD_OOB, &error_warn);
+    qemu_socket_select_nofail(sockfd, ioc->event,
+                              FD_READ | FD_ACCEPT | FD_CLOSE |
+                              FD_CONNECT | FD_WRITE | FD_OOB);
 
     source = g_source_new(&qio_channel_socket_source_funcs,
                           sizeof(QIOChannelSocketSource));
diff --git a/tests/unit/test-error-report.c b/tests/unit/test-error-report.c
index 54319c86c9..0cbde3c4cf 100644
--- a/tests/unit/test-error-report.c
+++ b/tests/unit/test-error-report.c
@@ -104,22 +104,6 @@ test_error_report_timestamp(void)
 ");
 }
 
-static void
-test_error_warn(void)
-{
-    if (g_test_subprocess()) {
-        error_setg(&error_warn, "Testing &error_warn");
-        return;
-    }
-
-    g_test_trap_subprocess(NULL, 0, 0);
-    g_test_trap_assert_passed();
-    g_test_trap_assert_stderr("\
-test-error-report: warning: Testing &error_warn*\
-");
-}
-
-
 int
 main(int argc, char *argv[])
 {
@@ -133,7 +117,6 @@ main(int argc, char *argv[])
     g_test_add_func("/error-report/glog", test_error_report_glog);
     g_test_add_func("/error-report/once", test_error_report_once);
     g_test_add_func("/error-report/timestamp", test_error_report_timestamp);
-    g_test_add_func("/error-report/warn", test_error_warn);
 
     return g_test_run();
 }
diff --git a/ui/gtk.c b/ui/gtk.c
index e91d093a49..9a08cadc88 100644
--- a/ui/gtk.c
+++ b/ui/gtk.c
@@ -1181,6 +1181,7 @@ static gboolean gd_touch_event(GtkWidget *widget, GdkEventTouch *touch,
                                void *opaque)
 {
     VirtualConsole *vc = opaque;
+    Error *err = NULL;
     uint64_t num_slot = GPOINTER_TO_UINT(touch->sequence);
     int type = -1;
 
@@ -1203,7 +1204,10 @@ static gboolean gd_touch_event(GtkWidget *widget, GdkEventTouch *touch,
     console_handle_touch_event(vc->gfx.dcl.con, touch_slots,
                                num_slot, surface_width(vc->gfx.ds),
                                surface_height(vc->gfx.ds), touch->x,
-                               touch->y, type, &error_warn);
+                               touch->y, type, &err);
+    if (err) {
+        warn_report_err(err);
+    }
     return TRUE;
 }
 
diff --git a/util/aio-win32.c b/util/aio-win32.c
index b125924433..c6fbce64c2 100644
--- a/util/aio-win32.c
+++ b/util/aio-win32.c
@@ -121,7 +121,7 @@ void aio_set_fd_handler(AioContext *ctx,
 
         QLIST_INSERT_HEAD_RCU(&ctx->aio_handlers, node, node);
         event = event_notifier_get_handle(&ctx->notifier);
-        qemu_socket_select(fd, event, bitmask, &error_warn);
+        qemu_socket_select_nofail(fd, event, bitmask);
     }
     if (old_node) {
         aio_remove_fd_handler(ctx, old_node);
diff --git a/util/error.c b/util/error.c
index daea2142f3..0ae08225c0 100644
--- a/util/error.c
+++ b/util/error.c
@@ -19,7 +19,6 @@
 
 Error *error_abort;
 Error *error_fatal;
-Error *error_warn;
 
 static void error_handle(Error **errp, Error *err)
 {
@@ -41,9 +40,7 @@ static void error_handle(Error **errp, Error *err)
         error_report_err(err);
         exit(1);
     }
-    if (errp == &error_warn) {
-        warn_report_err(err);
-    } else if (errp && !*errp) {
+    if (errp && !*errp) {
         *errp = err;
     } else {
         error_free(err);
diff --git a/util/oslib-win32.c b/util/oslib-win32.c
index df118b164e..e6109cc2cc 100644
--- a/util/oslib-win32.c
+++ b/util/oslib-win32.c
@@ -181,7 +181,7 @@ void qemu_socket_set_block(int fd)
 {
     unsigned long opt = 0;
 
-    qemu_socket_unselect(fd, &error_warn);
+    qemu_socket_unselect_nofail(fd);
     ioctlsocket(fd, FIONBIO, &opt);
 }
 
@@ -315,6 +315,25 @@ bool qemu_socket_unselect(int sockfd, Error **errp)
     return qemu_socket_select(sockfd, NULL, 0, errp);
 }
 
+void qemu_socket_select_nofail(int sockfd, WSAEVENT hEventObject,
+                               long lNetworkEvents)
+{
+    Error *err = NULL;
+
+    if (!qemu_socket_select(sockfd, hEventObject, lNetworkEvents, &err)) {
+        warn_report_err(err);
+    }
+}
+
+void qemu_socket_unselect_nofail(int sockfd)
+{
+    Error *err = NULL;
+
+    if (!qemu_socket_unselect(sockfd, &err)) {
+        warn_report_err(err);
+    }
+}
+
 int qemu_socketpair(int domain, int type, int protocol, int sv[2])
 {
     struct sockaddr_un addr = {
-- 
2.49.0

Re: [PATCH v2 12/12] error: Kill @error_warn

[Vladimir Sementsov-Ogievskiy]

On 17.09.25 14:52, Markus Armbruster wrote:
> We added @error_warn some two years ago in commit 3ffef1a55ca (error:
> add global &error_warn destination).  It has multiple issues:
> 
> * error.h's big comment was not updated for it.
> 
> * Function contracts were not updated for it.
> 
> * ERRP_GUARD() is unaware of @error_warn, and fails to mask it from
>    error_prepend() and such.  These crash on @error_warn, as pointed
>    out by Akihiko Odaki.
> 
> All fixable.  However, after more than two years, we had just of 15
> uses, of which the last few patches removed seven as unclean or
> otherwise undesirable, adding back five elsewhere.  I didn't look
> closely enough at the remaining seven to decide whether they are
> desirable or not.
> 
> I don't think this feature earns its keep.  Drop it.
> 
> Thanks-to: Akihiko  Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
> Signed-off-by: Markus Armbruster <armbru@redhat.com>

Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>


-- 
Best regards,
Vladimir