vl.c: call bdrv_init_with_whitelist() before cmdline parsing
PATCH 1/2 vl.c: call bdrv_init_with_whitelist() before cmdline parsing
commit 4d454574 "qemu-option: move standard option definitions out of qemu-config.c" broke support for commandline option groups that where registered during bdrv_init(). In particular support for -iscsi options was broken since that commit.
Fix by moving the bdrv_init_with_whitelist() before command line argument parsing.
--- vl.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
vl.c +2 -2
@@ -2935,6 +2935,8 @@ int main(int argc, char **argv, char **envp) nb_numa_nodes =0 ; nb_nics =0 ; + +bdrv_init_with_whitelist (); autostart=1 ;
@@ -4194,8 +4196,6 @@ int main(int argc, char **argv, char **envp)cpu_exec_init_all (); -bdrv_init_with_whitelist (); -blk_mig_init ();/* open the virtual block devices */
Signature · 1 line
1.7.9.5
PATCH 2/2 vl.c: throw an error if iscsi is not supported
this patch adds a check for qemu_find_opts("iscsi") returning NULL instead of blindly passing the result to qemu_opts_parse().
--- vl.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
vl.c +6 -3
@@ -3224,14 +3224,17 @@ int main(int argc, char **argv, char **envp)exit (1 ); }break ; -#ifdef CONFIG_LIBISCSI case QEMU_OPTION_iscsi: - opts =qemu_opts_parse (qemu_find_opts ("iscsi" ), optarg,0 ); + olist =qemu_find_opts ("iscsi" ); +if (!olist) { +fprintf (stderr,"iscsi is not supported by this qemu build.\n" ); +exit (1 ); + } + opts =qemu_opts_parse (olist, optarg,0 );if (!opts) {exit (1 ); }break ; -#endif #ifdef CONFIG_SLIRP case QEMU_OPTION_tftp: legacy_tftp_prefix = optarg;
Signature · 1 line
1.7.9.5
Re: [Qemu-devel] [PATCH 2/2] vl.c: throw an error if iscsi is not supported
Quoted 7 lines · Peter Lieven
this patch adds a check for
qemu_find_opts("iscsi")returning NULL instead of blindly passing the result toqemu_opts_parse().Signed-off-by Peter Lieven <[email protected]>--- vl.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
Why is this change necessary?
Stefan
Re: [Qemu-devel] [PATCH 2/2] vl.c: throw an error if iscsi is not supported
Am 25.03.2013 um 15:00 schrieb Stefan Hajnoczi <[email protected]>:
Quoted 10 lines · Peter Lieven
Quoted 7 lines
this patch adds a check for
qemu_find_opts("iscsi")returning NULL instead of blindly passing the result toqemu_opts_parse().Signed-off-by Peter Lieven <[email protected]>---
vl.c| 9 ++++++---1 file changed, 6 insertions(+), 3 deletions(-)Why is this change necessary?
a) just to make sure we check the result of qemu_find_opts for NULL in case a bug stops registering the opts again
b) to have an error output if libiscsi is not compiled in and someone passes an iscsi flag (analogue to the error if you supply -spice XXX)
Peter
Quoted 2 lines
Stefan
Re: [Qemu-devel] [PATCH 2/2] vl.c: throw an error if iscsi is not supported
Quoted 10 lines · Peter Lieven
Am 25.03.2013 um 15:00 schrieb Stefan Hajnoczi <[email protected]>:
Quoted 9 lines · Peter Lieven
Quoted 7 lines
this patch adds a check for
qemu_find_opts("iscsi")returning NULL instead of blindly passing the result toqemu_opts_parse().Signed-off-by Peter Lieven <[email protected]>---
vl.c| 9 ++++++---1 file changed, 6 insertions(+), 3 deletions(-)
Quoted 4 lines
Quoted 1 line
Why is this change necessary?
a) just to make sure we check the result of
qemu_find_optsfor NULL in case a bug stops registering the opts again
It's OK to crash for "can't happen" cases IMHO.
Quoted 2 lines
b) to have an error output if libiscsi is not compiled in and someone passes an iscsi flag (analogue to the error if you supply -spice XXX)
I don't have a strong opinion here but it would be good to be consistent. At the moment (as well as iscsi) SLIRP, TPM and mem_prealloc options all just vanish if qemu wasn't configured with them supported. [Various other things like SDL and Windows specific options do remain to produce an error.] So maybe we should move all these options to "always exist but may produce an error".
-- PMM
Re: [Qemu-devel] [PATCH 2/2] vl.c: throw an error if iscsi is not supported
Quoted 21 lines · Peter Maydell
Quoted 10 lines · Peter Lieven
Am 25.03.2013 um 15:00 schrieb Stefan Hajnoczi <[email protected]>:
Quoted 9 lines · Peter Lieven
Quoted 7 lines
this patch adds a check for
qemu_find_opts("iscsi")returning NULL instead of blindly passing the result toqemu_opts_parse().Signed-off-by Peter Lieven <[email protected]>---
vl.c| 9 ++++++---1 file changed, 6 insertions(+), 3 deletions(-)Quoted 4 lines
Quoted 1 line
Why is this change necessary?
a) just to make sure we check the result of
qemu_find_optsfor NULL in case a bug stops registering the opts againIt's OK to crash for "can't happen" cases IMHO.
Quoted 2 lines
b) to have an error output if libiscsi is not compiled in and someone passes an iscsi flag (analogue to the error if you supply -spice XXX)
Quoted 7 lines
I don't have a strong opinion here but it would be good to be consistent. At the moment (as well as iscsi) SLIRP, TPM and
mem_preallocoptions all just vanish if qemu wasn't configured with them supported. [Various other things like SDL and Windows specific options do remain to produce an error.] So maybe we should move all these options to "always exist but may produce an error".
Agreed. Either this patch should be dropped because it's not strictly necessary. Or we should be consistent and clean up all of vl.c:main(). There's not much point of just changing -iscsi.
Stefan
Stefan Hajnoczi
AppliedQuoted 5 lines · Peter Lieven
commit
4d454574"qemu-option: move standard option definitions out ofqemu-config.c" broke support for commandline option groups that where registered duringbdrv_init(). In particular support for -iscsi options was broken since that commit.
Quoted 7 lines
Fix by moving the
bdrv_init_with_whitelist()before command line argument parsing.Signed-off-by Peter Lieven <[email protected]>--- vl.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
Thanks, applied to my block tree:
https://github.com/stefanha/qemu/commits/block
vl.c int main(int argc, char **argv, char **envp) · line 2938
Quoted 9 lines
vl.c +1
-0@@ -2935,6 +2935,8 @@ int main(int argc, char **argv, char **envp) nb_numa_nodes =0 ; nb_nics =0 ; +
I removed the trailing whitespace here. git-am(1) warns about it.