| Message ID | 1490621195-2228-9-git-send-email-armbru@redhat.com |
|---|---|
| State | New |
| Headers | show |
On 27.03.2017 15:26, Markus Armbruster wrote: > This reverts half of commit 0a55679. We're having second thoughts on > the QAPI schema (and thus the external interface), and haven't reached > consensus, yet. Issues include: > > * The implementation uses deprecated rados_conf_set() key > "auth_supported". No biggie. > > * The implementation makes -drive silently ignore invalid parameters > "auth" and "auth-supported.*.X" where X isn't "auth". Fixable (in > fact I'm going to fix similar bugs around parameter server), so > again no biggie. > > * BlockdevOptionsRbd member @password-secret applies only to > authentication method cephx. Should it be a variant member of > RbdAuthMethod? > > * BlockdevOptionsRbd member @user could apply to both methods cephx > and none, but I'm not sure it's actually used with none. If it > isn't, should it be a variant member of RbdAuthMethod? > > * The client offers a *set* of authentication methods, not a list. > Should the methods be optional members of BlockdevOptionsRbd instead > of members of list @auth-supported? The latter begs the question > what multiple entries for the same method mean. Trivial question > now that RbdAuthMethod contains nothing but @type, but less so when > RbdAuthMethod acquires other members, such the ones discussed above. > > * How BlockdevOptionsRbd member @auth-supported interacts with > settings from a configuration file specified with @conf is > undocumented. I suspect it's untested, too. > > Let's avoid painting ourselves into a corner now, and revert the > feature for 2.9. > > Note that users can still configure authentication methods with a > configuration file. They probably do that anyway if they use Ceph > outside QEMU as well. > > qemu_rbd_array_opts()'s parameter @type now must be RBD_MON_HOST, > which is silly. This will be cleaned up shortly. > > Signed-off-by: Markus Armbruster <armbru@redhat.com> > --- > block/rbd.c | 31 +++---------------------------- > qapi/block-core.json | 24 ------------------------ > 2 files changed, 3 insertions(+), 52 deletions(-) Reviewed-by: Max Reitz <mreitz@redhat.com>
On 03/27/2017 08:26 AM, Markus Armbruster wrote: > This reverts half of commit 0a55679. We're having second thoughts on > the QAPI schema (and thus the external interface), and haven't reached > consensus, yet. Issues include: > > Let's avoid painting ourselves into a corner now, and revert the > feature for 2.9. > > Note that users can still configure authentication methods with a > configuration file. They probably do that anyway if they use Ceph > outside QEMU as well. If we're only reverting the QMP blockdev-add feature, then this makes absolute sense (it's not a regression since we don't have a release with it yet, and we don't want to bake something into the release that can't be supported). But breaking -drive usage seems risky, especially since libvirt is already expecting to work - I'm worried that doing this may break existing libvirt command line usage if the QemuOpts side doesn't permit anything at all. Maybe we need to rely on your '=foo' or 'x-foo' hack for letting QemuOpts still accept the old spelling during -drive but not during QMP. > > qemu_rbd_array_opts()'s parameter @type now must be RBD_MON_HOST, > which is silly. This will be cleaned up shortly. > > Signed-off-by: Markus Armbruster <armbru@redhat.com> > --- > block/rbd.c | 31 +++---------------------------- > qapi/block-core.json | 24 ------------------------ > 2 files changed, 3 insertions(+), 52 deletions(-) > > +++ b/qapi/block-core.json > @@ -2601,27 +2601,6 @@ > > > ## > -# @RbdAuthSupport: > -# > -# An enumeration of RBD auth support > -# > -# Since: 2.9 > -## > -{ 'enum': 'RbdAuthSupport', > - 'data': [ 'cephx', 'none' ] } > - > - > -## > -# @RbdAuthMethod: > -# > -# An enumeration of rados auth_supported types > -# > -# Since: 2.9 > -## > -{ 'struct': 'RbdAuthMethod', > - 'data': { 'auth': 'RbdAuthSupport' } } > - Removing the .json QMP support is fine. But I'm reluctant to give R-b without knowing for sure that -drive usage won't regress.
Eric Blake <eblake@redhat.com> writes: > On 03/27/2017 08:26 AM, Markus Armbruster wrote: >> This reverts half of commit 0a55679. We're having second thoughts on >> the QAPI schema (and thus the external interface), and haven't reached >> consensus, yet. Issues include: >> > >> Let's avoid painting ourselves into a corner now, and revert the >> feature for 2.9. >> >> Note that users can still configure authentication methods with a >> configuration file. They probably do that anyway if they use Ceph >> outside QEMU as well. > > If we're only reverting the QMP blockdev-add feature, then this makes > absolute sense (it's not a regression since we don't have a release with > it yet, and we don't want to bake something into the release that can't > be supported). But breaking -drive usage seems risky, especially since > libvirt is already expecting to work - I'm worried that doing this may > break existing libvirt command line usage if the QemuOpts side doesn't > permit anything at all. Maybe we need to rely on your '=foo' or 'x-foo' > hack for letting QemuOpts still accept the old spelling during -drive > but not during QMP. > >> >> qemu_rbd_array_opts()'s parameter @type now must be RBD_MON_HOST, >> which is silly. This will be cleaned up shortly. >> >> Signed-off-by: Markus Armbruster <armbru@redhat.com> >> --- >> block/rbd.c | 31 +++---------------------------- >> qapi/block-core.json | 24 ------------------------ >> 2 files changed, 3 insertions(+), 52 deletions(-) >> > >> +++ b/qapi/block-core.json >> @@ -2601,27 +2601,6 @@ >> >> >> ## >> -# @RbdAuthSupport: >> -# >> -# An enumeration of RBD auth support >> -# >> -# Since: 2.9 >> -## >> -{ 'enum': 'RbdAuthSupport', >> - 'data': [ 'cephx', 'none' ] } >> - >> - >> -## >> -# @RbdAuthMethod: >> -# >> -# An enumeration of rados auth_supported types >> -# >> -# Since: 2.9 >> -## >> -{ 'struct': 'RbdAuthMethod', >> - 'data': { 'auth': 'RbdAuthSupport' } } >> - > > Removing the .json QMP support is fine. But I'm reluctant to give R-b > without knowing for sure that -drive usage won't regress. auth-supported landed in master only on March 2nd. What libvirt usage exactly do you think this could break?
On 03/27/2017 01:31 PM, Markus Armbruster wrote: >>> - >>> -## >>> -# @RbdAuthMethod: >>> -# >>> -# An enumeration of rados auth_supported types >>> -# >>> -# Since: 2.9 >>> -## >>> -{ 'struct': 'RbdAuthMethod', >>> - 'data': { 'auth': 'RbdAuthSupport' } } >>> - >> >> Removing the .json QMP support is fine. But I'm reluctant to give R-b >> without knowing for sure that -drive usage won't regress. > > auth-supported landed in master only on March 2nd. auth-supported via -blockdev-add only landed on March 2nd. But auth-supported via -drive landed in commit 60390a2, Jan 2016; and is in use by libvirt: src/qemu/qemu_command.c: ":key=%s:auth_supported=cephx\\;none", > > What libvirt usage exactly do you think this could break? > Libvirt has been managing rbd drives using -drive since at least libvirt commit 5745cd1, in Nov 2011, where even back then it was passing: virBufferEscape(opt, ":", ":key=%s:auth_supported=cephx none", (back when it used space instead of ; to separate the list of supported auth types). As I've never personally used RBD (whether through qemu directly, or through libvirt), I'm extremely wary of breaking -drive usage that "works" (for some definition of "works"), even though I have no qualms making the QMP interface extremely limited.
Eric Blake <eblake@redhat.com> writes: > On 03/27/2017 01:31 PM, Markus Armbruster wrote: > >>>> - >>>> -## >>>> -# @RbdAuthMethod: >>>> -# >>>> -# An enumeration of rados auth_supported types >>>> -# >>>> -# Since: 2.9 >>>> -## >>>> -{ 'struct': 'RbdAuthMethod', >>>> - 'data': { 'auth': 'RbdAuthSupport' } } >>>> - >>> >>> Removing the .json QMP support is fine. But I'm reluctant to give R-b >>> without knowing for sure that -drive usage won't regress. >> >> auth-supported landed in master only on March 2nd. > > auth-supported via -blockdev-add only landed on March 2nd. But > auth-supported via -drive landed in commit 60390a2, Jan 2016; and is in > use by libvirt: > > src/qemu/qemu_command.c: > ":key=%s:auth_supported=cephx\\;none", That's a key-value part of the pseudo-filename. *Not* reverted by this patch. Only QemuOpts parameter auth_supported is. QemuOpts parameter: -drive driver=rbd,auth_supported.0.auth=none,... Pseudo-filename: -drive file=rbd:...:auth_supported=none >> What libvirt usage exactly do you think this could break? >> > > Libvirt has been managing rbd drives using -drive since at least libvirt > commit 5745cd1, in Nov 2011, where even back then it was passing: > virBufferEscape(opt, ":", ":key=%s:auth_supported=cephx none", > > (back when it used space instead of ; to separate the list of supported > auth types). > > As I've never personally used RBD (whether through qemu directly, or > through libvirt), I'm extremely wary of breaking -drive usage that > "works" (for some definition of "works"), even though I have no qualms > making the QMP interface extremely limited. All clear now?
On 03/27/2017 02:14 PM, Markus Armbruster wrote: >>>> >>>> Removing the .json QMP support is fine. But I'm reluctant to give R-b >>>> without knowing for sure that -drive usage won't regress. >>> >>> auth-supported landed in master only on March 2nd. >> >> auth-supported via -blockdev-add only landed on March 2nd. But >> auth-supported via -drive landed in commit 60390a2, Jan 2016; and is in >> use by libvirt: >> >> src/qemu/qemu_command.c: >> ":key=%s:auth_supported=cephx\\;none", > > That's a key-value part of the pseudo-filename. *Not* reverted by this > patch. Only QemuOpts parameter auth_supported is. > > QemuOpts parameter: -drive driver=rbd,auth_supported.0.auth=none,... > Pseudo-filename: -drive file=rbd:...:auth_supported=none > >>> What libvirt usage exactly do you think this could break? >>> >> >> Libvirt has been managing rbd drives using -drive since at least libvirt >> commit 5745cd1, in Nov 2011, where even back then it was passing: >> virBufferEscape(opt, ":", ":key=%s:auth_supported=cephx none", >> >> (back when it used space instead of ; to separate the list of supported >> auth types). Ah, so as long as libvirt uses 'drive file=rbd:...' with key-value pairs, our backdoor of =key-values will let it continue to work. And looking more at libvirt, it definitely looks like it is sticking to file=rbd: pseudo-filenames for now; for example: tests/qemuxml2argvdata/qemuxml2argv-disk-drive-network-rbd-auth-AES.args:-drive 'file=rbd:pool/image:id=myname:auth_supported=cephx\;none:\ I'm not sure if (or where) a big fat comment to that effect would be beneficial, but it certainly goes a long ways in explaining the goals of your series. >> >> As I've never personally used RBD (whether through qemu directly, or >> through libvirt), I'm extremely wary of breaking -drive usage that >> "works" (for some definition of "works"), even though I have no qualms >> making the QMP interface extremely limited. > > All clear now? I think so. Commit 60390a2 (Jan 2016) mentions the use of auth_supported in the commit message (matching libvirt usage), but under the pseudo-file format; the actual rbd.c file did not special case it differently from any other key-value pair. "auth" wasn't added to QemuOpts (for the -drive driver=rbd, form) until commit 0a55679, which is unreleased, so libvirt can't have been relying on it, and ripping it out now is safe enough. Under that light, I'm now going to re-read your patch...
On 03/27/2017 08:26 AM, Markus Armbruster wrote: > This reverts half of commit 0a55679. We're having second thoughts on > the QAPI schema (and thus the external interface), and haven't reached > consensus, yet. Issues include: > > Let's avoid painting ourselves into a corner now, and revert the > feature for 2.9. There may still be some tweaks to improve the commit message and/or code comments to clarify things that tripped me up until later in the subthread, but now that I understand the difference between pseudo-file format (where the key-value pair backdoor still works for libvirt's usage of -drive file=rbd:...) and QemuOpts format (-drive driver=rbd,... which didn't really exist in 2.8, and where we don't want to bake in something we don't like in 2.9), I agree with the move. > > Note that users can still configure authentication methods with a > configuration file. They probably do that anyway if they use Ceph > outside QEMU as well. > > qemu_rbd_array_opts()'s parameter @type now must be RBD_MON_HOST, > which is silly. This will be cleaned up shortly. > > Signed-off-by: Markus Armbruster <armbru@redhat.com> > --- > block/rbd.c | 31 +++---------------------------- > qapi/block-core.json | 24 ------------------------ > 2 files changed, 3 insertions(+), 52 deletions(-) > Reviewed-by: Eric Blake <eblake@redhat.com>
On Mon, Mar 27, 2017 at 03:26:32PM +0200, Markus Armbruster wrote: > This reverts half of commit 0a55679. We're having second thoughts on > the QAPI schema (and thus the external interface), and haven't reached > consensus, yet. Issues include: > > * The implementation uses deprecated rados_conf_set() key > "auth_supported". No biggie. > > * The implementation makes -drive silently ignore invalid parameters > "auth" and "auth-supported.*.X" where X isn't "auth". Fixable (in > fact I'm going to fix similar bugs around parameter server), so > again no biggie. > > * BlockdevOptionsRbd member @password-secret applies only to > authentication method cephx. Should it be a variant member of > RbdAuthMethod? > > * BlockdevOptionsRbd member @user could apply to both methods cephx > and none, but I'm not sure it's actually used with none. If it > isn't, should it be a variant member of RbdAuthMethod? > > * The client offers a *set* of authentication methods, not a list. > Should the methods be optional members of BlockdevOptionsRbd instead > of members of list @auth-supported? The latter begs the question > what multiple entries for the same method mean. Trivial question > now that RbdAuthMethod contains nothing but @type, but less so when > RbdAuthMethod acquires other members, such the ones discussed above. > > * How BlockdevOptionsRbd member @auth-supported interacts with > settings from a configuration file specified with @conf is > undocumented. I suspect it's untested, too. > > Let's avoid painting ourselves into a corner now, and revert the > feature for 2.9. > > Note that users can still configure authentication methods with a > configuration file. They probably do that anyway if they use Ceph > outside QEMU as well. > > qemu_rbd_array_opts()'s parameter @type now must be RBD_MON_HOST, > which is silly. This will be cleaned up shortly. > > Signed-off-by: Markus Armbruster <armbru@redhat.com> I think this move makes sense; it allows blockdev-add to still be supported for rbd, but does not lock us into a perhaps unwieldy API. Reviewed-by: Jeff Cody <jcody@redhat.com> > --- > block/rbd.c | 31 +++---------------------------- > qapi/block-core.json | 24 ------------------------ > 2 files changed, 3 insertions(+), 52 deletions(-) > > diff --git a/block/rbd.c b/block/rbd.c > index cf0bab0..103ce44 100644 > --- a/block/rbd.c > +++ b/block/rbd.c > @@ -320,8 +320,7 @@ static QemuOptsList runtime_opts = { > .help = "Rados id name", > }, > /* > - * server.* and auth-supported.* extracted manually, see > - * qemu_rbd_array_opts() > + * server.* extracted manually, see qemu_rbd_array_opts() > */ > { > .name = "password-secret", > @@ -356,11 +355,6 @@ static QemuOptsList runtime_opts = { > .name = "port", > .type = QEMU_OPT_STRING, > }, > - { > - .name = "auth", > - .type = QEMU_OPT_STRING, > - .help = "Supported authentication method, either cephx or none", > - }, > { /* end of list */ } > }, > }; > @@ -512,7 +506,6 @@ static void qemu_rbd_complete_aio(RADOSCB *rcb) > } > > #define RBD_MON_HOST 0 > -#define RBD_AUTH_SUPPORTED 1 > > static char *qemu_rbd_array_opts(QDict *options, const char *prefix, int type, > Error **errp) > @@ -527,7 +520,7 @@ static char *qemu_rbd_array_opts(QDict *options, const char *prefix, int type, > Error *local_err = NULL; > int i; > > - assert(type == RBD_MON_HOST || type == RBD_AUTH_SUPPORTED); > + assert(type == RBD_MON_HOST); > > num_entries = qdict_array_entries(options, prefix); > > @@ -573,10 +566,9 @@ static char *qemu_rbd_array_opts(QDict *options, const char *prefix, int type, > value = strbuf; > } > } else { > - value = qemu_opt_get(opts, "auth"); > + abort(); > } > > - > /* each iteration in the for loop will build upon the string, and if > * rados_str is NULL then it is our first pass */ > if (rados_str) { > @@ -608,7 +600,6 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags, > QemuOpts *opts; > Error *local_err = NULL; > char *mon_host = NULL; > - char *auth_supported = NULL; > int r; > > opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort); > @@ -619,14 +610,6 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags, > return -EINVAL; > } > > - auth_supported = qemu_rbd_array_opts(options, "auth-supported.", > - RBD_AUTH_SUPPORTED, &local_err); > - if (local_err) { > - error_propagate(errp, local_err); > - r = -EINVAL; > - goto failed_opts; > - } > - > mon_host = qemu_rbd_array_opts(options, "server.", > RBD_MON_HOST, &local_err); > if (local_err) { > @@ -678,13 +661,6 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags, > } > } > > - if (auth_supported) { > - r = rados_conf_set(s->cluster, "auth_supported", auth_supported); > - if (r < 0) { > - goto failed_shutdown; > - } > - } > - > if (qemu_rbd_set_auth(s->cluster, secretid, errp) < 0) { > r = -EIO; > goto failed_shutdown; > @@ -735,7 +711,6 @@ failed_shutdown: > failed_opts: > qemu_opts_del(opts); > g_free(mon_host); > - g_free(auth_supported); > return r; > } > > diff --git a/qapi/block-core.json b/qapi/block-core.json > index 5d2efe4..6a7ca0b 100644 > --- a/qapi/block-core.json > +++ b/qapi/block-core.json > @@ -2601,27 +2601,6 @@ > > > ## > -# @RbdAuthSupport: > -# > -# An enumeration of RBD auth support > -# > -# Since: 2.9 > -## > -{ 'enum': 'RbdAuthSupport', > - 'data': [ 'cephx', 'none' ] } > - > - > -## > -# @RbdAuthMethod: > -# > -# An enumeration of rados auth_supported types > -# > -# Since: 2.9 > -## > -{ 'struct': 'RbdAuthMethod', > - 'data': { 'auth': 'RbdAuthSupport' } } > - > -## > # @BlockdevOptionsRbd: > # > # @pool: Ceph pool name. > @@ -2639,8 +2618,6 @@ > # @server: Monitor host address and port. This maps > # to the "mon_host" Ceph option. > # > -# @auth-supported: Authentication supported. > -# > # @password-secret: The ID of a QCryptoSecret object providing > # the password for the login. > # > @@ -2653,7 +2630,6 @@ > '*snapshot': 'str', > '*user': 'str', > '*server': ['InetSocketAddressBase'], > - '*auth-supported': ['RbdAuthMethod'], > '*password-secret': 'str' } } > > ## > -- > 2.7.4 >
Eric Blake <eblake@redhat.com> writes: > On 03/27/2017 08:26 AM, Markus Armbruster wrote: >> This reverts half of commit 0a55679. We're having second thoughts on >> the QAPI schema (and thus the external interface), and haven't reached >> consensus, yet. Issues include: >> > >> Let's avoid painting ourselves into a corner now, and revert the >> feature for 2.9. > > There may still be some tweaks to improve the commit message and/or code > comments to clarify things that tripped me up until later in the > subthread, but now that I understand the difference between pseudo-file > format (where the key-value pair backdoor still works for libvirt's > usage of -drive file=rbd:...) and QemuOpts format (-drive driver=rbd,... > which didn't really exist in 2.8, and where we don't want to bake in > something we don't like in 2.9), I agree with the move. I'm inserting ... >> Note that users can still configure authentication methods with a >> configuration file. They probably do that anyway if they use Ceph >> outside QEMU as well. ... this hint right here: Further note that this doesn't affect use of key "auth-supported" in -drive file=rbd:...:key=value. Good enough? >> qemu_rbd_array_opts()'s parameter @type now must be RBD_MON_HOST, >> which is silly. This will be cleaned up shortly. >> >> Signed-off-by: Markus Armbruster <armbru@redhat.com> >> --- >> block/rbd.c | 31 +++---------------------------- >> qapi/block-core.json | 24 ------------------------ >> 2 files changed, 3 insertions(+), 52 deletions(-) >> > > Reviewed-by: Eric Blake <eblake@redhat.com> Thanks!
On 03/28/2017 03:24 AM, Markus Armbruster wrote: > Eric Blake <eblake@redhat.com> writes: > >> On 03/27/2017 08:26 AM, Markus Armbruster wrote: >>> This reverts half of commit 0a55679. We're having second thoughts on >>> the QAPI schema (and thus the external interface), and haven't reached >>> consensus, yet. Issues include: >>> >> >>> Let's avoid painting ourselves into a corner now, and revert the >>> feature for 2.9. >> >> There may still be some tweaks to improve the commit message and/or code >> comments to clarify things that tripped me up until later in the >> subthread, but now that I understand the difference between pseudo-file >> format (where the key-value pair backdoor still works for libvirt's >> usage of -drive file=rbd:...) and QemuOpts format (-drive driver=rbd,... >> which didn't really exist in 2.8, and where we don't want to bake in >> something we don't like in 2.9), I agree with the move. > > I'm inserting ... > >>> Note that users can still configure authentication methods with a >>> configuration file. They probably do that anyway if they use Ceph >>> outside QEMU as well. > > ... this hint right here: > > Further note that this doesn't affect use of key "auth-supported" in > -drive file=rbd:...:key=value. > > Good enough? Yes, that does the trick.
diff --git a/block/rbd.c b/block/rbd.c index cf0bab0..103ce44 100644 --- a/block/rbd.c +++ b/block/rbd.c @@ -320,8 +320,7 @@ static QemuOptsList runtime_opts = { .help = "Rados id name", }, /* - * server.* and auth-supported.* extracted manually, see - * qemu_rbd_array_opts() + * server.* extracted manually, see qemu_rbd_array_opts() */ { .name = "password-secret", @@ -356,11 +355,6 @@ static QemuOptsList runtime_opts = { .name = "port", .type = QEMU_OPT_STRING, }, - { - .name = "auth", - .type = QEMU_OPT_STRING, - .help = "Supported authentication method, either cephx or none", - }, { /* end of list */ } }, }; @@ -512,7 +506,6 @@ static void qemu_rbd_complete_aio(RADOSCB *rcb) } #define RBD_MON_HOST 0 -#define RBD_AUTH_SUPPORTED 1 static char *qemu_rbd_array_opts(QDict *options, const char *prefix, int type, Error **errp) @@ -527,7 +520,7 @@ static char *qemu_rbd_array_opts(QDict *options, const char *prefix, int type, Error *local_err = NULL; int i; - assert(type == RBD_MON_HOST || type == RBD_AUTH_SUPPORTED); + assert(type == RBD_MON_HOST); num_entries = qdict_array_entries(options, prefix); @@ -573,10 +566,9 @@ static char *qemu_rbd_array_opts(QDict *options, const char *prefix, int type, value = strbuf; } } else { - value = qemu_opt_get(opts, "auth"); + abort(); } - /* each iteration in the for loop will build upon the string, and if * rados_str is NULL then it is our first pass */ if (rados_str) { @@ -608,7 +600,6 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags, QemuOpts *opts; Error *local_err = NULL; char *mon_host = NULL; - char *auth_supported = NULL; int r; opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort); @@ -619,14 +610,6 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags, return -EINVAL; } - auth_supported = qemu_rbd_array_opts(options, "auth-supported.", - RBD_AUTH_SUPPORTED, &local_err); - if (local_err) { - error_propagate(errp, local_err); - r = -EINVAL; - goto failed_opts; - } - mon_host = qemu_rbd_array_opts(options, "server.", RBD_MON_HOST, &local_err); if (local_err) { @@ -678,13 +661,6 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags, } } - if (auth_supported) { - r = rados_conf_set(s->cluster, "auth_supported", auth_supported); - if (r < 0) { - goto failed_shutdown; - } - } - if (qemu_rbd_set_auth(s->cluster, secretid, errp) < 0) { r = -EIO; goto failed_shutdown; @@ -735,7 +711,6 @@ failed_shutdown: failed_opts: qemu_opts_del(opts); g_free(mon_host); - g_free(auth_supported); return r; } diff --git a/qapi/block-core.json b/qapi/block-core.json index 5d2efe4..6a7ca0b 100644 --- a/qapi/block-core.json +++ b/qapi/block-core.json @@ -2601,27 +2601,6 @@ ## -# @RbdAuthSupport: -# -# An enumeration of RBD auth support -# -# Since: 2.9 -## -{ 'enum': 'RbdAuthSupport', - 'data': [ 'cephx', 'none' ] } - - -## -# @RbdAuthMethod: -# -# An enumeration of rados auth_supported types -# -# Since: 2.9 -## -{ 'struct': 'RbdAuthMethod', - 'data': { 'auth': 'RbdAuthSupport' } } - -## # @BlockdevOptionsRbd: # # @pool: Ceph pool name. @@ -2639,8 +2618,6 @@ # @server: Monitor host address and port. This maps # to the "mon_host" Ceph option. # -# @auth-supported: Authentication supported. -# # @password-secret: The ID of a QCryptoSecret object providing # the password for the login. # @@ -2653,7 +2630,6 @@ '*snapshot': 'str', '*user': 'str', '*server': ['InetSocketAddressBase'], - '*auth-supported': ['RbdAuthMethod'], '*password-secret': 'str' } } ##
This reverts half of commit 0a55679. We're having second thoughts on the QAPI schema (and thus the external interface), and haven't reached consensus, yet. Issues include: * The implementation uses deprecated rados_conf_set() key "auth_supported". No biggie. * The implementation makes -drive silently ignore invalid parameters "auth" and "auth-supported.*.X" where X isn't "auth". Fixable (in fact I'm going to fix similar bugs around parameter server), so again no biggie. * BlockdevOptionsRbd member @password-secret applies only to authentication method cephx. Should it be a variant member of RbdAuthMethod? * BlockdevOptionsRbd member @user could apply to both methods cephx and none, but I'm not sure it's actually used with none. If it isn't, should it be a variant member of RbdAuthMethod? * The client offers a *set* of authentication methods, not a list. Should the methods be optional members of BlockdevOptionsRbd instead of members of list @auth-supported? The latter begs the question what multiple entries for the same method mean. Trivial question now that RbdAuthMethod contains nothing but @type, but less so when RbdAuthMethod acquires other members, such the ones discussed above. * How BlockdevOptionsRbd member @auth-supported interacts with settings from a configuration file specified with @conf is undocumented. I suspect it's untested, too. Let's avoid painting ourselves into a corner now, and revert the feature for 2.9. Note that users can still configure authentication methods with a configuration file. They probably do that anyway if they use Ceph outside QEMU as well. qemu_rbd_array_opts()'s parameter @type now must be RBD_MON_HOST, which is silly. This will be cleaned up shortly. Signed-off-by: Markus Armbruster <armbru@redhat.com> --- block/rbd.c | 31 +++---------------------------- qapi/block-core.json | 24 ------------------------ 2 files changed, 3 insertions(+), 52 deletions(-)