| Message ID | 1338815410-24890-2-git-send-email-coreyb@linux.vnet.ibm.com |
|---|---|
| State | New |
| Headers | show |
Am 04.06.2012 15:10, schrieb Corey Bryant: > This patch adds QMP support for the getfd command using the QAPI framework. > Like the HMP getfd command, it is used to pass a file descriptor via > SCM_RIGHTS. However, the QMP getfd command also returns the received file > descriptor, which is a difference in behavior from the HMP getfd command, > which returns nothing. > > Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com> > --- > hmp-commands.hx | 2 +- > monitor.c | 37 ++++++++++++++++++++++++++++++++++++- > qapi-schema.json | 13 +++++++++++++ > qmp-commands.hx | 6 ++++-- > 4 files changed, 54 insertions(+), 4 deletions(-) > > diff --git a/hmp-commands.hx b/hmp-commands.hx > index 18cb415..dfab369 100644 > --- a/hmp-commands.hx > +++ b/hmp-commands.hx > @@ -1211,7 +1211,7 @@ ETEXI > .params = "getfd name", > .help = "receive a file descriptor via SCM rights and assign it a name", > .user_print = monitor_user_noop, > - .mhandler.cmd_new = do_getfd, > + .mhandler.cmd_new = hmp_getfd, > }, > > STEXI > diff --git a/monitor.c b/monitor.c > index 12a6fe2..6acf5a3 100644 > --- a/monitor.c > +++ b/monitor.c > @@ -2199,7 +2199,7 @@ static void do_inject_mce(Monitor *mon, const QDict *qdict) > } > #endif > > -static int do_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) > +static int hmp_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) > { > const char *fdname = qdict_get_str(qdict, "fdname"); > mon_fd_t *monfd; This should become a wrapper around qmp_getfd() instead of duplicating the logic. Kevin
On 06/04/2012 10:45 AM, Kevin Wolf wrote: > Am 04.06.2012 15:10, schrieb Corey Bryant: >> This patch adds QMP support for the getfd command using the QAPI framework. >> Like the HMP getfd command, it is used to pass a file descriptor via >> SCM_RIGHTS. However, the QMP getfd command also returns the received file >> descriptor, which is a difference in behavior from the HMP getfd command, >> which returns nothing. >> >> Signed-off-by: Corey Bryant<coreyb@linux.vnet.ibm.com> >> --- >> hmp-commands.hx | 2 +- >> monitor.c | 37 ++++++++++++++++++++++++++++++++++++- >> qapi-schema.json | 13 +++++++++++++ >> qmp-commands.hx | 6 ++++-- >> 4 files changed, 54 insertions(+), 4 deletions(-) >> >> diff --git a/hmp-commands.hx b/hmp-commands.hx >> index 18cb415..dfab369 100644 >> --- a/hmp-commands.hx >> +++ b/hmp-commands.hx >> @@ -1211,7 +1211,7 @@ ETEXI >> .params = "getfd name", >> .help = "receive a file descriptor via SCM rights and assign it a name", >> .user_print = monitor_user_noop, >> - .mhandler.cmd_new = do_getfd, >> + .mhandler.cmd_new = hmp_getfd, >> }, >> >> STEXI >> diff --git a/monitor.c b/monitor.c >> index 12a6fe2..6acf5a3 100644 >> --- a/monitor.c >> +++ b/monitor.c >> @@ -2199,7 +2199,7 @@ static void do_inject_mce(Monitor *mon, const QDict *qdict) >> } >> #endif >> >> -static int do_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) >> +static int hmp_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) >> { >> const char *fdname = qdict_get_str(qdict, "fdname"); >> mon_fd_t *monfd; > > This should become a wrapper around qmp_getfd() instead of duplicating > the logic. > > Kevin > Yes, I was thinking about that. I'll do this in v2.
On Mon, 4 Jun 2012 09:10:08 -0400 Corey Bryant <coreyb@linux.vnet.ibm.com> wrote: > This patch adds QMP support for the getfd command using the QAPI framework. > Like the HMP getfd command, it is used to pass a file descriptor via > SCM_RIGHTS. However, the QMP getfd command also returns the received file > descriptor, which is a difference in behavior from the HMP getfd command, > which returns nothing. I have a few comments regarding the qapi conversion below, but something important to discuss is that returning an int the way you're doing it is certainly incompatible. Today, we return a dict on success: { "return": {} } But this patch changes it to: { "return": 42 } There are two ways to do this without breaking compatibility: 1. Add a new command (say get-file-descriptor) 2. Return a type instead, like: { "return": { "file-descriptor": 42 } } I think I prefer item 1, as we could also take the opportunity to fix the argument type and improve its name. Besides, we don't have a schema to do 2. Also, please split the getfd conversion to the qapi from changing its return type in two patches (even if we go for a new command, please do convert getfd too, for completeness); and also convert closefd, please. More comments below. > Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com> > --- > hmp-commands.hx | 2 +- > monitor.c | 37 ++++++++++++++++++++++++++++++++++++- > qapi-schema.json | 13 +++++++++++++ > qmp-commands.hx | 6 ++++-- > 4 files changed, 54 insertions(+), 4 deletions(-) > > diff --git a/hmp-commands.hx b/hmp-commands.hx > index 18cb415..dfab369 100644 > --- a/hmp-commands.hx > +++ b/hmp-commands.hx > @@ -1211,7 +1211,7 @@ ETEXI > .params = "getfd name", > .help = "receive a file descriptor via SCM rights and assign it a name", > .user_print = monitor_user_noop, The .user_print line should be dropped. > - .mhandler.cmd_new = do_getfd, > + .mhandler.cmd_new = hmp_getfd, You should use 'cmd' (instead of 'cmd_new'). Please, look for other qapi conversions for examples (like a15fef21c). > }, > > STEXI > diff --git a/monitor.c b/monitor.c > index 12a6fe2..6acf5a3 100644 > --- a/monitor.c > +++ b/monitor.c > @@ -2199,7 +2199,7 @@ static void do_inject_mce(Monitor *mon, const QDict *qdict) > } > #endif > > -static int do_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) > +static int hmp_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) This should be moved to hmp.c and the function signature should be different, looking at other qapi conversions will help you. > { > const char *fdname = qdict_get_str(qdict, "fdname"); > mon_fd_t *monfd; > @@ -2235,6 +2235,41 @@ static int do_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) > return 0; > } > > +int64_t qmp_getfd(const char *fdname, Error **errp) > +{ > + mon_fd_t *monfd; > + int fd; > + > + fd = qemu_chr_fe_get_msgfd(cur_mon->chr); > + if (fd == -1) { > + error_set(errp, QERR_FD_NOT_SUPPLIED); > + return -1; > + } > + > + if (qemu_isdigit(fdname[0])) { > + error_set(errp, QERR_INVALID_PARAMETER_VALUE, "fdname", > + "a name not starting with a digit"); > + return -1; > + } > + > + QLIST_FOREACH(monfd, &cur_mon->fds, next) { > + if (strcmp(monfd->name, fdname) != 0) { > + continue; > + } > + > + close(monfd->fd); > + monfd->fd = fd; > + return fd; > + } > + > + monfd = g_malloc0(sizeof(mon_fd_t)); > + monfd->name = g_strdup(fdname); > + monfd->fd = fd; > + > + QLIST_INSERT_HEAD(&cur_mon->fds, monfd, next); > + return fd; > +} > + > static int do_closefd(Monitor *mon, const QDict *qdict, QObject **ret_data) > { > const char *fdname = qdict_get_str(qdict, "fdname"); > diff --git a/qapi-schema.json b/qapi-schema.json > index 2ca7195..5f26ba2 100644 > --- a/qapi-schema.json > +++ b/qapi-schema.json > @@ -1755,3 +1755,16 @@ > # Since: 0.14.0 > ## > { 'command': 'device_del', 'data': {'id': 'str'} } > + > +## > +# @getfd: > +# > +# Receive a file descriptor via SCM rights and assign it a name > +# > +# @fdname: file descriptor name > +# > +# Returns: The QEMU @fd that was received > +# > +# Since: 1.2 This command exists since 0.14 if I'm not mistaken, only the return type is new. > +## > +{ 'command': 'getfd', 'data': {'fdname': 'str'}, 'returns': 'int' } > diff --git a/qmp-commands.hx b/qmp-commands.hx > index db980fa..e13d583 100644 > --- a/qmp-commands.hx > +++ b/qmp-commands.hx > @@ -844,7 +844,7 @@ EQMP > .params = "getfd name", > .help = "receive a file descriptor via SCM rights and assign it a name", > .user_print = monitor_user_noop, > - .mhandler.cmd_new = do_getfd, > + .mhandler.cmd_new = qmp_marshal_input_getfd, > }, > > SQMP > @@ -857,10 +857,12 @@ Arguments: > > - "fdname": file descriptor name (json-string) > > +Return a json-int with the QEMU file descriptor that was received. > + > Example: > > -> { "execute": "getfd", "arguments": { "fdname": "fd1" } } > -<- { "return": {} } > +<- { "return": 42 } > > EQMP >
On 06/05/2012 02:30 PM, Luiz Capitulino wrote: > On Mon, 4 Jun 2012 09:10:08 -0400 > Corey Bryant<coreyb@linux.vnet.ibm.com> wrote: > >> This patch adds QMP support for the getfd command using the QAPI framework. >> Like the HMP getfd command, it is used to pass a file descriptor via >> SCM_RIGHTS. However, the QMP getfd command also returns the received file >> descriptor, which is a difference in behavior from the HMP getfd command, >> which returns nothing. > > I have a few comments regarding the qapi conversion below, but something > important to discuss is that returning an int the way you're doing it is > certainly incompatible. Thanks for your feedback. > > Today, we return a dict on success: > > { "return": {} } > > But this patch changes it to: > > { "return": 42 } > > There are two ways to do this without breaking compatibility: > > 1. Add a new command (say get-file-descriptor) What do you think about using getfd2 for the command name? I'm thinking getfd2 may be more obvious that it corresponds to closefd. That assumes we'll use the same array internally to store fds and closefd can be used to close the fd opened by get-file-descriptor/getfd2. I assume this approach would still return an int: { "return": 42 } > > 2. Return a type instead, like: > > { "return": { "file-descriptor": 42 } } > > I think I prefer item 1, as we could also take the opportunity to fix the > argument type and improve its name. Besides, we don't have a schema to do 2. Is it fdname that you think could be improved? fdname seems pretty straight forward to me. > > Also, please split the getfd conversion to the qapi from changing its > return type in two patches (even if we go for a new command, please do > convert getfd too, for completeness); and also convert closefd, please. Ok > > More comments below. > >> Signed-off-by: Corey Bryant<coreyb@linux.vnet.ibm.com> >> --- >> hmp-commands.hx | 2 +- >> monitor.c | 37 ++++++++++++++++++++++++++++++++++++- >> qapi-schema.json | 13 +++++++++++++ >> qmp-commands.hx | 6 ++++-- >> 4 files changed, 54 insertions(+), 4 deletions(-) >> >> diff --git a/hmp-commands.hx b/hmp-commands.hx >> index 18cb415..dfab369 100644 >> --- a/hmp-commands.hx >> +++ b/hmp-commands.hx >> @@ -1211,7 +1211,7 @@ ETEXI >> .params = "getfd name", >> .help = "receive a file descriptor via SCM rights and assign it a name", >> .user_print = monitor_user_noop, > > The .user_print line should be dropped. Ok > >> - .mhandler.cmd_new = do_getfd, >> + .mhandler.cmd_new = hmp_getfd, > > You should use 'cmd' (instead of 'cmd_new'). Please, look for other qapi > conversions for examples (like a15fef21c). Ok > >> }, >> >> STEXI >> diff --git a/monitor.c b/monitor.c >> index 12a6fe2..6acf5a3 100644 >> --- a/monitor.c >> +++ b/monitor.c >> @@ -2199,7 +2199,7 @@ static void do_inject_mce(Monitor *mon, const QDict *qdict) >> } >> #endif >> >> -static int do_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) >> +static int hmp_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) > > This should be moved to hmp.c and the function signature should be different, > looking at other qapi conversions will help you. Ok > >> { >> const char *fdname = qdict_get_str(qdict, "fdname"); >> mon_fd_t *monfd; >> @@ -2235,6 +2235,41 @@ static int do_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) >> return 0; >> } >> >> +int64_t qmp_getfd(const char *fdname, Error **errp) >> +{ >> + mon_fd_t *monfd; >> + int fd; >> + >> + fd = qemu_chr_fe_get_msgfd(cur_mon->chr); >> + if (fd == -1) { >> + error_set(errp, QERR_FD_NOT_SUPPLIED); >> + return -1; >> + } >> + >> + if (qemu_isdigit(fdname[0])) { >> + error_set(errp, QERR_INVALID_PARAMETER_VALUE, "fdname", >> + "a name not starting with a digit"); >> + return -1; >> + } >> + >> + QLIST_FOREACH(monfd,&cur_mon->fds, next) { >> + if (strcmp(monfd->name, fdname) != 0) { >> + continue; >> + } >> + >> + close(monfd->fd); >> + monfd->fd = fd; >> + return fd; >> + } >> + >> + monfd = g_malloc0(sizeof(mon_fd_t)); >> + monfd->name = g_strdup(fdname); >> + monfd->fd = fd; >> + >> + QLIST_INSERT_HEAD(&cur_mon->fds, monfd, next); >> + return fd; >> +} >> + >> static int do_closefd(Monitor *mon, const QDict *qdict, QObject **ret_data) >> { >> const char *fdname = qdict_get_str(qdict, "fdname"); >> diff --git a/qapi-schema.json b/qapi-schema.json >> index 2ca7195..5f26ba2 100644 >> --- a/qapi-schema.json >> +++ b/qapi-schema.json >> @@ -1755,3 +1755,16 @@ >> # Since: 0.14.0 >> ## >> { 'command': 'device_del', 'data': {'id': 'str'} } >> + >> +## >> +# @getfd: >> +# >> +# Receive a file descriptor via SCM rights and assign it a name >> +# >> +# @fdname: file descriptor name >> +# >> +# Returns: The QEMU @fd that was received >> +# >> +# Since: 1.2 > > This command exists since 0.14 if I'm not mistaken, only the return type > is new. Ok > >> +## >> +{ 'command': 'getfd', 'data': {'fdname': 'str'}, 'returns': 'int' } >> diff --git a/qmp-commands.hx b/qmp-commands.hx >> index db980fa..e13d583 100644 >> --- a/qmp-commands.hx >> +++ b/qmp-commands.hx >> @@ -844,7 +844,7 @@ EQMP >> .params = "getfd name", >> .help = "receive a file descriptor via SCM rights and assign it a name", >> .user_print = monitor_user_noop, >> - .mhandler.cmd_new = do_getfd, >> + .mhandler.cmd_new = qmp_marshal_input_getfd, >> }, >> >> SQMP >> @@ -857,10 +857,12 @@ Arguments: >> >> - "fdname": file descriptor name (json-string) >> >> +Return a json-int with the QEMU file descriptor that was received. >> + >> Example: >> >> -> { "execute": "getfd", "arguments": { "fdname": "fd1" } } >> -<- { "return": {} } >> +<- { "return": 42 } >> >> EQMP >> > >
On Wed, 06 Jun 2012 10:04:23 -0400 Corey Bryant <coreyb@linux.vnet.ibm.com> wrote: > > Today, we return a dict on success: > > > > { "return": {} } > > > > But this patch changes it to: > > > > { "return": 42 } > > > > There are two ways to do this without breaking compatibility: > > > > 1. Add a new command (say get-file-descriptor) > > What do you think about using getfd2 for the command name? I'm thinking > getfd2 may be more obvious that it corresponds to closefd. We're going for more descriptive names in QMP. I don't have strong objections against get-fd2 if there's consensus that 'fd' is better than 'file-descriptor', although 'fd2' is a bit confusing. > That assumes > we'll use the same array internally to store fds and closefd can be used > to close the fd opened by get-file-descriptor/getfd2. You mean using the same array for getfd and get-file-descriptor? Yes, the descriptor list is global. > I assume this approach would still return an int: { "return": 42 } The new command? Yes. > > 2. Return a type instead, like: > > > > { "return": { "file-descriptor": 42 } } > > > > I think I prefer item 1, as we could also take the opportunity to fix the > > argument type and improve its name. Besides, we don't have a schema to do 2. > > Is it fdname that you think could be improved? fdname seems pretty > straight forward to me. What I'm trying to avoid is having too short names when that's not necessary. I think I'd just use 'name' or 'file-descriptor-name' for the verbose option, but I don't have strong objections against 'fdname'.
On 06/06/2012 01:50 PM, Luiz Capitulino wrote: > On Wed, 06 Jun 2012 10:04:23 -0400 > Corey Bryant<coreyb@linux.vnet.ibm.com> wrote: > >>> Today, we return a dict on success: >>> >>> { "return": {} } >>> >>> But this patch changes it to: >>> >>> { "return": 42 } >>> >>> There are two ways to do this without breaking compatibility: >>> >>> 1. Add a new command (say get-file-descriptor) >> >> What do you think about using getfd2 for the command name? I'm thinking >> getfd2 may be more obvious that it corresponds to closefd. > > We're going for more descriptive names in QMP. I don't have strong objections > against get-fd2 if there's consensus that 'fd' is better than 'file-descriptor', > although 'fd2' is a bit confusing. I really don't have a strong opinion either so I'll do whatever the consensus wants. Does anyone else have an opinion? > >> That assumes >> we'll use the same array internally to store fds and closefd can be used >> to close the fd opened by get-file-descriptor/getfd2. > > You mean using the same array for getfd and get-file-descriptor? Yes, the > descriptor list is global. > Yes, that's what I meant. >> I assume this approach would still return an int: { "return": 42 } > > The new command? Yes. > Ok >>> 2. Return a type instead, like: >>> >>> { "return": { "file-descriptor": 42 } } >>> >>> I think I prefer item 1, as we could also take the opportunity to fix the >>> argument type and improve its name. Besides, we don't have a schema to do 2. >> >> Is it fdname that you think could be improved? fdname seems pretty >> straight forward to me. > > What I'm trying to avoid is having too short names when that's not necessary. > I think I'd just use 'name' or 'file-descriptor-name' for the verbose option, > but I don't have strong objections against 'fdname'. > Unless anyone objects I'll plan on going with more descriptive names, since that is the desired direction for QMP. So we'll have get-file-descriptor for the new command name and file-descriptor-name for the argument name.
On Wed, Jun 06, 2012 at 10:04:23AM -0400, Corey Bryant wrote: > > > On 06/05/2012 02:30 PM, Luiz Capitulino wrote: > >On Mon, 4 Jun 2012 09:10:08 -0400 > >Corey Bryant<coreyb@linux.vnet.ibm.com> wrote: > > > >>This patch adds QMP support for the getfd command using the QAPI framework. > >>Like the HMP getfd command, it is used to pass a file descriptor via > >>SCM_RIGHTS. However, the QMP getfd command also returns the received file > >>descriptor, which is a difference in behavior from the HMP getfd command, > >>which returns nothing. > > > >I have a few comments regarding the qapi conversion below, but something > >important to discuss is that returning an int the way you're doing it is > >certainly incompatible. > > Thanks for your feedback. > > > > >Today, we return a dict on success: > > > > { "return": {} } > > > >But this patch changes it to: > > > > { "return": 42 } > > > >There are two ways to do this without breaking compatibility: > > > > 1. Add a new command (say get-file-descriptor) > > What do you think about using getfd2 for the command name? I'm > thinking getfd2 may be more obvious that it corresponds to closefd. > That assumes we'll use the same array internally to store fds and > closefd can be used to close the fd opened by > get-file-descriptor/getfd2. How about calling it 'passfd' instead ? I have always thought the name 'getfd' to be a little wierd really, since we're not getting an FD from QEMU, we're passing it one that we have. Regards, Daniel
On 06/08/2012 06:46 AM, Daniel P. Berrange wrote: > On Wed, Jun 06, 2012 at 10:04:23AM -0400, Corey Bryant wrote: >> >> >> On 06/05/2012 02:30 PM, Luiz Capitulino wrote: >>> On Mon, 4 Jun 2012 09:10:08 -0400 >>> Corey Bryant<coreyb@linux.vnet.ibm.com> wrote: >>> >>>> This patch adds QMP support for the getfd command using the QAPI framework. >>>> Like the HMP getfd command, it is used to pass a file descriptor via >>>> SCM_RIGHTS. However, the QMP getfd command also returns the received file >>>> descriptor, which is a difference in behavior from the HMP getfd command, >>>> which returns nothing. >>> >>> I have a few comments regarding the qapi conversion below, but something >>> important to discuss is that returning an int the way you're doing it is >>> certainly incompatible. >> >> Thanks for your feedback. >> >>> >>> Today, we return a dict on success: >>> >>> { "return": {} } >>> >>> But this patch changes it to: >>> >>> { "return": 42 } >>> >>> There are two ways to do this without breaking compatibility: >>> >>> 1. Add a new command (say get-file-descriptor) >> >> What do you think about using getfd2 for the command name? I'm >> thinking getfd2 may be more obvious that it corresponds to closefd. >> That assumes we'll use the same array internally to store fds and >> closefd can be used to close the fd opened by >> get-file-descriptor/getfd2. > > How about calling it 'passfd' instead ? I have always thought > the name 'getfd' to be a little wierd really, since we're not > getting an FD from QEMU, we're passing it one that we have. I agree, thanks for the suggestion. I know Luiz suggested detailed command names, and I think this will be ok and self documents itself fairly well with the compact 'passfd'. I have v2 patches that are just about ready, so I'll update them to use 'passfd'.
diff --git a/hmp-commands.hx b/hmp-commands.hx index 18cb415..dfab369 100644 --- a/hmp-commands.hx +++ b/hmp-commands.hx @@ -1211,7 +1211,7 @@ ETEXI .params = "getfd name", .help = "receive a file descriptor via SCM rights and assign it a name", .user_print = monitor_user_noop, - .mhandler.cmd_new = do_getfd, + .mhandler.cmd_new = hmp_getfd, }, STEXI diff --git a/monitor.c b/monitor.c index 12a6fe2..6acf5a3 100644 --- a/monitor.c +++ b/monitor.c @@ -2199,7 +2199,7 @@ static void do_inject_mce(Monitor *mon, const QDict *qdict) } #endif -static int do_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) +static int hmp_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) { const char *fdname = qdict_get_str(qdict, "fdname"); mon_fd_t *monfd; @@ -2235,6 +2235,41 @@ static int do_getfd(Monitor *mon, const QDict *qdict, QObject **ret_data) return 0; } +int64_t qmp_getfd(const char *fdname, Error **errp) +{ + mon_fd_t *monfd; + int fd; + + fd = qemu_chr_fe_get_msgfd(cur_mon->chr); + if (fd == -1) { + error_set(errp, QERR_FD_NOT_SUPPLIED); + return -1; + } + + if (qemu_isdigit(fdname[0])) { + error_set(errp, QERR_INVALID_PARAMETER_VALUE, "fdname", + "a name not starting with a digit"); + return -1; + } + + QLIST_FOREACH(monfd, &cur_mon->fds, next) { + if (strcmp(monfd->name, fdname) != 0) { + continue; + } + + close(monfd->fd); + monfd->fd = fd; + return fd; + } + + monfd = g_malloc0(sizeof(mon_fd_t)); + monfd->name = g_strdup(fdname); + monfd->fd = fd; + + QLIST_INSERT_HEAD(&cur_mon->fds, monfd, next); + return fd; +} + static int do_closefd(Monitor *mon, const QDict *qdict, QObject **ret_data) { const char *fdname = qdict_get_str(qdict, "fdname"); diff --git a/qapi-schema.json b/qapi-schema.json index 2ca7195..5f26ba2 100644 --- a/qapi-schema.json +++ b/qapi-schema.json @@ -1755,3 +1755,16 @@ # Since: 0.14.0 ## { 'command': 'device_del', 'data': {'id': 'str'} } + +## +# @getfd: +# +# Receive a file descriptor via SCM rights and assign it a name +# +# @fdname: file descriptor name +# +# Returns: The QEMU @fd that was received +# +# Since: 1.2 +## +{ 'command': 'getfd', 'data': {'fdname': 'str'}, 'returns': 'int' } diff --git a/qmp-commands.hx b/qmp-commands.hx index db980fa..e13d583 100644 --- a/qmp-commands.hx +++ b/qmp-commands.hx @@ -844,7 +844,7 @@ EQMP .params = "getfd name", .help = "receive a file descriptor via SCM rights and assign it a name", .user_print = monitor_user_noop, - .mhandler.cmd_new = do_getfd, + .mhandler.cmd_new = qmp_marshal_input_getfd, }, SQMP @@ -857,10 +857,12 @@ Arguments: - "fdname": file descriptor name (json-string) +Return a json-int with the QEMU file descriptor that was received. + Example: -> { "execute": "getfd", "arguments": { "fdname": "fd1" } } -<- { "return": {} } +<- { "return": 42 } EQMP
This patch adds QMP support for the getfd command using the QAPI framework. Like the HMP getfd command, it is used to pass a file descriptor via SCM_RIGHTS. However, the QMP getfd command also returns the received file descriptor, which is a difference in behavior from the HMP getfd command, which returns nothing. Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com> --- hmp-commands.hx | 2 +- monitor.c | 37 ++++++++++++++++++++++++++++++++++++- qapi-schema.json | 13 +++++++++++++ qmp-commands.hx | 6 ++++-- 4 files changed, 54 insertions(+), 4 deletions(-)