Bug: Sandbox: libvirt breakdowns qemu guest
v2 cover Bug: Sandbox: libvirt breakdowns qemu guest
1. Problem Description
======================
If QEMU is built without seccomp support, 'elevateprivileges' remains compiled. This option of sandbox is treated as an indication for seccomp blacklist support in libvirt. This behavior is introduced by the libvirt commits 31ca6a5 and 3527f9d. It would make libvirt build wrong QEMU cmdline, and then the guest startup would fail.
2. Libvirt Log
==============
qemu-system-s390x: -sandbox on,obsolete=deny,elevateprivileges=deny,spawn=deny,\
resourcecontrol=deny: seccomp support is disabled
3. Fixup
========
Compile the code related to sandbox only when CONFIG_SECCOMP is defined.
Yi Min Zhao (1):
sandbox: disable -sandbox if CONFIG_SECCOMP undefined
vl.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)
Signature · 1 line
Yi Min
PATCH v2 1/1 sandbox: disable -sandbox if CONFIG_SECCOMP undefined
If CONFIG_SECCOMP is undefined, the option 'elevateprivileges' remains compiled. This would make libvirt set the corresponding capability and then trigger the guest startup fails. So this patch excludes the code regarding seccomp staff if CONFIG_SECCOMP is undefined.
--- vl.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)
vl.c +8 -5
@@ -257,6 +257,7 @@ static QemuOptsList qemu_rtc_opts = { }, }; +#ifdef CONFIG_SECCOMP static QemuOptsList qemu_sandbox_opts = { .name ="sandbox" , .implied_opt_name = "enable",
@@ -285,6 +286,7 @@ static QemuOptsList qemu_sandbox_opts = { {/* end of list */ } }, }; +#endif static QemuOptsList qemu_option_rom_opts = { .name ="option-rom" ,
@@ -1041,10 +1043,10 @@ static int bt_parse(const char *opt)return 1 ; } +#ifdef CONFIG_SECCOMP static int parse_sandbox (void *opaque,QemuOpts *opts,Error **errp) {if (qemu_opt_get_bool (opts,"enable" , false)) { -#ifdef CONFIG_SECCOMP uint32_t seccomp_opts =QEMU_SECCOMP_SET_DEFAULT |QEMU_SECCOMP_SET_OBSOLETE ;const char *value =NULL ;
@@ -1114,14 +1116,11 @@ static int parse_sandbox(void *opaque, QemuOpts *opts, Error **errp) "in the kernel"); return -1; } -#else -error_report ("seccomp support is disabled" ); -return -1 ; -#endif }return 0 ; } +#endif static int parse_name (void *opaque,QemuOpts *opts,Error **errp) {
@@ -3074,7 +3073,9 @@ int main(int argc, char **argv, char **envp)qemu_add_opts (&qemu_mem_opts);qemu_add_opts (&qemu_smp_opts);qemu_add_opts (&qemu_boot_opts); +#ifdef CONFIG_SECCOMP qemu_add_opts (&qemu_sandbox_opts); +#endif qemu_add_opts (&qemu_add_fd_opts);qemu_add_opts (&qemu_object_opts);qemu_add_opts (&qemu_tpmdev_opts);
@@ -4071,10 +4072,12 @@ int main(int argc, char **argv, char **envp)exit (1 ); } +#ifdef CONFIG_SECCOMP if (qemu_opts_foreach (qemu_find_opts ("sandbox" ), parse_sandbox,NULL ,NULL )) {exit (1 ); } +#endif if (qemu_opts_foreach (qemu_find_opts ("name" ), parse_name,NULL ,NULL )) {
Signature · 1 line
Yi Min
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 3 lines · Yi Min Zhao
If
CONFIG_SECCOMPis undefined, the option 'elevateprivileges' remains compiled. This would make libvirt set the corresponding capability and then trigger the guest startup fails. So this patch excludes the code
s/trigger the guest startup fails/trigger failure during guest startup/
Quoted 1 line
regarding seccomp staff if
CONFIG_SECCOMPis undefined.
s/staff/command line options/
Quoted 6 lines
Signed-off-by Yi Min Zhao <[email protected]>--- vl.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)
A maintainer can touch up the commit message, so:
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 12 lines · Eric Blake
Quoted 3 lines · Yi Min Zhao
If
CONFIG_SECCOMPis undefined, the option 'elevateprivileges' remains compiled. This would make libvirt set the corresponding capability and then trigger the guest startup fails. So this patch excludes the codes/trigger the guest startup fails/trigger failure during guest startup/
Quoted 1 line
regarding seccomp staff if
CONFIG_SECCOMPis undefined.s/staff/command line options/
Quoted 1 line
Quoted 9 lines
Quoted 5 lines
Signed-off-by Yi Min Zhao <[email protected]>---
vl.c| 13 ++++++++-----1 file changed, 8 insertions(+), 5 deletions(-)A maintainer can touch up the commit message, so:
Reviewed-by Eric Blake <[email protected]>
Thanks for your comments! Have updated commit msg.
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
In reply to Eric BlakeAdd Paolo to CC list. @Paolo, expect your comment. Thanks very much!
Quoted 21 lines · Eric Blake
Quoted 3 lines · Yi Min Zhao
If
CONFIG_SECCOMPis undefined, the option 'elevateprivileges' remains compiled. This would make libvirt set the corresponding capability and then trigger the guest startup fails. So this patch excludes the codes/trigger the guest startup fails/trigger failure during guest startup/
Quoted 1 line
regarding seccomp staff if
CONFIG_SECCOMPis undefined.s/staff/command line options/
Quoted 6 lines
Signed-off-by Yi Min Zhao <[email protected]>---
vl.c| 13 ++++++++-----1 file changed, 8 insertions(+), 5 deletions(-)A maintainer can touch up the commit message, so:
Reviewed-by Eric Blake <[email protected]>
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
In reply to Yi Min ZhaoQuoted 4 lines · Yi Min Zhao
If
CONFIG_SECCOMPis undefined, the option 'elevateprivileges' remains compiled. This would make libvirt set the corresponding capability and then trigger the guest startup fails. So this patch excludes the code regarding seccomp staff ifCONFIG_SECCOMPis undefined.
Just a sugestion for the next patch you send: If it's a single patch, you don't need to format it with a cover-letter. Just put all the description in the body, or if you need to add a text that shouldn't be included in the commit message, just add it after the "---" after Signed-off-by.
Quoted 39 lines
Signed-off-by Yi Min Zhao <[email protected]>--- vl.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)vl.c +3
-1@@ -257,6 +257,7 @@ static QemuOptsList qemu_rtc_opts = { }, }; +#ifdef CONFIG_SECCOMP static QemuOptsList qemu_sandbox_opts = { .name ="sandbox" , .implied_opt_name = "enable",@@ -285,6 +286,7 @@ static QemuOptsList qemu_sandbox_opts = { {/* end of list */ } }, }; +#endif static QemuOptsList qemu_option_rom_opts = { .name ="option-rom" ,@@ -1041,10 +1043,10 @@ static int bt_parse(const char *opt)return 1 ; } +#ifdef CONFIG_SECCOMP static int parse_sandbox (void *opaque,QemuOpts *opts,Error **errp) {if (qemu_opt_get_bool (opts,"enable" , false)) { -#ifdef CONFIG_SECCOMP uint32_t seccomp_opts =QEMU_SECCOMP_SET_DEFAULT |QEMU_SECCOMP_SET_OBSOLETE ;const char *value =NULL ;
vl.c static int parse_sandbox(void *opaque, QemuOpts *opts, Error **errp) · line 1120
Quoted 7 lines
vl.c +0
-4@@ -1114,14 +1116,11 @@ static int parse_sandbox(void *opaque, QemuOpts *opts, Error **errp) "in the kernel");return -1 ; } -#else -error_report ("seccomp support is disabled" ); -return -1 ; -#endif
Any reason not to keep the error message on the new #endif location?
Quoted 34 lines
vl.c +5
-0}return 0 ; } +#endif static int parse_name (void *opaque,QemuOpts *opts,Error **errp) {@@ -3074,7 +3073,9 @@ int main(int argc, char **argv, char **envp)qemu_add_opts (&qemu_mem_opts);qemu_add_opts (&qemu_smp_opts);qemu_add_opts (&qemu_boot_opts); +#ifdef CONFIG_SECCOMP qemu_add_opts (&qemu_sandbox_opts); +#endif qemu_add_opts (&qemu_add_fd_opts);qemu_add_opts (&qemu_object_opts);qemu_add_opts (&qemu_tpmdev_opts);@@ -4071,10 +4072,12 @@ int main(int argc, char **argv, char **envp)exit (1 ); } +#ifdef CONFIG_SECCOMP if (qemu_opts_foreach (qemu_find_opts ("sandbox" ), parse_sandbox,NULL ,NULL )) {exit (1 ); } +#endif if (qemu_opts_foreach (qemu_find_opts ("name" ), parse_name,NULL ,NULL )) {Signature · 1 line
Yi Min
I just wanted a review from Ján, since he is the author of the original libvirt patch. Does this breaks libvirt logic in any way? If not, ACK on this patch.
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 9 lines · Eduardo Otubo
Quoted 4 lines · Yi Min Zhao
If
CONFIG_SECCOMPis undefined, the option 'elevateprivileges' remains compiled. This would make libvirt set the corresponding capability and then trigger the guest startup fails. So this patch excludes the code regarding seccomp staff ifCONFIG_SECCOMPis undefined.Just a sugestion for the next patch you send: If it's a single patch, you don't need to format it with a cover-letter. Just put all the description in the body, or if you need to add a text that shouldn't be included in the commit message, just add it after the "---" after Signed-off-by.
OK. Thanks for your suggestion.
Quoted 39 lines
Quoted 39 lines
Signed-off-by Yi Min Zhao <[email protected]>--- vl.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)vl.c +3
-1@@ -257,6 +257,7 @@ static QemuOptsList qemu_rtc_opts = { }, }; +#ifdef CONFIG_SECCOMP static QemuOptsList qemu_sandbox_opts = { .name ="sandbox" , .implied_opt_name = "enable",@@ -285,6 +286,7 @@ static QemuOptsList qemu_sandbox_opts = { {/* end of list */ } }, }; +#endif static QemuOptsList qemu_option_rom_opts = { .name ="option-rom" ,@@ -1041,10 +1043,10 @@ static int bt_parse(const char *opt)return 1 ; } +#ifdef CONFIG_SECCOMP static int parse_sandbox (void *opaque,QemuOpts *opts,Error **errp) {if (qemu_opt_get_bool (opts,"enable" , false)) { -#ifdef CONFIG_SECCOMP uint32_t seccomp_opts =QEMU_SECCOMP_SET_DEFAULT |QEMU_SECCOMP_SET_OBSOLETE ;const char *value =NULL ;@@ -1114,14 +1116,11 @@ static int parse_sandbox(void *opaque, QemuOpts *opts, Error **errp)
Quoted 8 lines
vl.c static int parse_sandbox(void *opaque, QemuOpts *opts, Error **errp) · line 1120
Quoted 7 lines
vl.c +0
-4"in the kernel");return -1 ; } -#else -error_report ("seccomp support is disabled" ); -return -1 ; -#endif Any reason not to keep the error message on the new #endif location?
If error report is originally wrapped in CONFIG_SECCOMP undefined. This patch excludes the entire function if CONFIG_SECCOMP is undefined. So the error report is not needed.
Quoted 30 lines
Quoted 30 lines
vl.c +5
-0}return 0 ; } +#endif static int parse_name (void *opaque,QemuOpts *opts,Error **errp) {@@ -3074,7 +3073,9 @@ int main(int argc, char **argv, char **envp)qemu_add_opts (&qemu_mem_opts);qemu_add_opts (&qemu_smp_opts);qemu_add_opts (&qemu_boot_opts); +#ifdef CONFIG_SECCOMP qemu_add_opts (&qemu_sandbox_opts); +#endif qemu_add_opts (&qemu_add_fd_opts);qemu_add_opts (&qemu_object_opts);qemu_add_opts (&qemu_tpmdev_opts);@@ -4071,10 +4072,12 @@ int main(int argc, char **argv, char **envp)exit (1 ); } +#ifdef CONFIG_SECCOMP if (qemu_opts_foreach (qemu_find_opts ("sandbox" ), parse_sandbox,NULL ,NULL )) {exit (1 ); } +#endif
Quoted 9 lines
Quoted 5 lines
vl.c +0
-0if (qemu_opts_foreach (qemu_find_opts ("name" ), parse_name,NULL ,NULL )) {Signature · 1 line
Yi MinI just wanted a review from Ján, since he is the author of the original libvirt patch. Does this breaks libvirt logic in any way? If not, ACK on this patch.
OK.
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
In reply to Eduardo OtuboQuoted 29 lines · Eduardo Otubo
Quoted 4 lines · Yi Min Zhao
If
CONFIG_SECCOMPis undefined, the option 'elevateprivileges' remains compiled. This would make libvirt set the corresponding capability and then trigger the guest startup fails. So this patch excludes the code regarding seccomp staff ifCONFIG_SECCOMPis undefined.Just a sugestion for the next patch you send: If it's a single patch, you don't need to format it with a cover-letter. Just put all the description in the body, or if you need to add a text that shouldn't be included in the commit message, just add it after the "---" after Signed-off-by.
Quoted 18 lines
Signed-off-by Yi Min Zhao <[email protected]>--- vl.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)vl.c +2
-0@@ -4071,10 +4072,12 @@ int main(int argc, char **argv, char **envp)exit (1 ); } +#ifdef CONFIG_SECCOMP if (qemu_opts_foreach (qemu_find_opts ("sandbox" ), parse_sandbox,NULL ,NULL )) {exit (1 ); } +#endif
Quoted 9 lines
Quoted 5 lines
vl.c +0
-1if (qemu_opts_foreach (qemu_find_opts ("name" ), parse_name,NULL ,NULL )) { --Yi Min
I just wanted a review from Ján, since he is the author of the original libvirt patch. Does this breaks libvirt logic in any way? If not, ACK on this patch.
Current libvirt logic assumes the -sandbox option is always present. (IIRC it was introduced in QEMU 1.1 and when we switched from help scraping to capability probing via QMP for QEMU 1.2, there was no way to detect it)
This patch fixes the usage of QEMU new enough for seccomp blacklist (where libvirt enables the sandbox by default),
but breaks the usage of QEMU with compiled out sandbox and setting
seccomp_sandbox = 0
in libvirt's qemu.conf:
error: internal error: process exited while connecting to monitor:
qemu-git: -sandbox off: There is no option group 'sandbox'
But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make seccomp_sandbox = 0 a no-op if it's compiled out.
Jano
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 52 lines · Ján Tomko
Quoted 29 lines · Eduardo Otubo
Quoted 4 lines · Yi Min Zhao
If
CONFIG_SECCOMPis undefined, the option 'elevateprivileges' remains compiled. This would make libvirt set the corresponding capability and then trigger the guest startup fails. So this patch excludes the code regarding seccomp staff ifCONFIG_SECCOMPis undefined.Just a sugestion for the next patch you send: If it's a single patch, you don't need to format it with a cover-letter. Just put all the description in the body, or if you need to add a text that shouldn't be included in the commit message, just add it after the "---" after Signed-off-by.
Quoted 18 lines
Signed-off-by Yi Min Zhao <[email protected]>--- vl.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)vl.c +2
-0@@ -4071,10 +4072,12 @@ int main(int argc, char **argv, char **envp)exit (1 ); } +#ifdef CONFIG_SECCOMP if (qemu_opts_foreach (qemu_find_opts ("sandbox" ), parse_sandbox,NULL ,NULL )) {exit (1 ); } +#endif Quoted 9 lines
Quoted 5 lines
vl.c +0
-1if (qemu_opts_foreach (qemu_find_opts ("name" ), parse_name,NULL ,NULL )) { --Yi Min
I just wanted a review from Ján, since he is the author of the original libvirt patch. Does this breaks libvirt logic in any way? If not, ACK on this patch.
Current libvirt logic assumes the -sandbox option is always present. (IIRC it was introduced in QEMU 1.1 and when we switched from help scraping to capability probing via QMP for QEMU 1.2, there was no way to detect it)
This patch fixes the usage of QEMU new enough for seccomp blacklist (where libvirt enables the sandbox by default),
but breaks the usage of QEMU with compiled out sandbox and setting
seccomp_sandbox= 0
in libvirt's qemu.conf:
Quoted 8 lines
error: internal error: process exited while connecting to monitor:
qemu-git: -sandbox off: There is no option group 'sandbox'But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make
seccomp_sandbox= 0 a no-op if it's compiled out.
This looks like a good solution for the libvirt side. Can you add this support so we can merge this fix?
Thanks a lot,
Signature · 1 line
Eduardo Otubo
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 20 lines · Eduardo Otubo
Quoted 19 lines · Ján Tomko
Quoted 17 lines · Eduardo Otubo
Quoted 4 lines · Yi Min Zhao
If
CONFIG_SECCOMPis undefined, the option 'elevateprivileges' remains compiled. This would make libvirt set the corresponding capability and then trigger the guest startup fails. So this patch excludes the code regarding seccomp staff ifCONFIG_SECCOMPis undefined.Just a sugestion for the next patch you send: If it's a single patch, you don't need to format it with a cover-letter. Just put all the description in the body, or if you need to add a text that shouldn't be included in the commit message, just add it after the "---" after Signed-off-by.
Quoted 6 lines
Signed-off-by Yi Min Zhao <[email protected]>--- vl.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)
[...]
Quoted 16 lines
Quoted 16 lines
Current libvirt logic assumes the -sandbox option is always present. (IIRC it was introduced in QEMU 1.1 and when we switched from help scraping to capability probing via QMP for QEMU 1.2, there was no way to detect it)
This patch fixes the usage of QEMU new enough for seccomp blacklist (where libvirt enables the sandbox by default),
but breaks the usage of QEMU with compiled out sandbox and setting
seccomp_sandbox= 0
in libvirt's qemu.conf:error: internal error: process exited while connecting to monitor:
qemu-git: -sandbox off: There is no option group 'sandbox'
Quoted 8 lines
Quoted 4 lines
But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make
seccomp_sandbox= 0 a no-op if it's compiled out.This looks like a good solution for the libvirt side. Can you add this support so we can merge this fix?
Patches proposed:
https://www.redhat.com/archives/libvir-list/2018-May/msg01430.html
Jano
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 51 lines · Ján Tomko
Quoted 26 lines · Eduardo Otubo
Quoted 25 lines · Ján Tomko
Quoted 2 lines · Eduardo Otubo
Quoted 1 line · Yi Min Zhao
If
CONFIG_SECCOMPis undefined, the option 'elevateprivileges'remains
Quoted 1 line
Quoted 1 line
compiled. This would make libvirt set the corresponding
capability and
Quoted 1 line
Quoted 1 line
then trigger the guest startup fails. So this patch excludes the
code
Quoted 3 lines
Quoted 1 line
regarding seccomp staff if
CONFIG_SECCOMPis undefined.Just a sugestion for the next patch you send: If it's a single
patch, you don't
Quoted 1 line
need to format it with a cover-letter. Just put all the
description in the body,
Quoted 1 line
or if you need to add a text that shouldn't be included in the
commit message,
Quoted 8 lines
just add it after the "---" after Signed-off-by.
Quoted 6 lines
Signed-off-by Yi Min Zhao <[email protected]>---
vl.c| 13 ++++++++-----1 file changed, 8 insertions(+), 5 deletions(-)[...]
Quoted 21 lines
Quoted 20 lines
Current libvirt logic assumes the -sandbox option is always present. (IIRC it was introduced in QEMU 1.1 and when we switched from help scraping to capability probing via QMP for QEMU 1.2, there was no way to detect it)
This patch fixes the usage of QEMU new enough for seccomp blacklist (where libvirt enables the sandbox by default),
but breaks the usage of QEMU with compiled out sandbox and setting
seccomp_sandbox = 0
in libvirt's qemu.conf:error: internal error: process exited while connecting to monitor:
qemu-git: -sandbox off: There is no option group 'sandbox'But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make
seccomp_sandbox= 0 a no-op if it's compiled out.
Quoted 8 lines
Quoted 3 lines
This looks like a good solution for the libvirt side. Can you add this support so we can merge this fix?
Patches proposed:
https://www.redhat.com/archives/libvir-list/2018-May/msg01430.htmlJano
Thanks for your work!
Ján Tomko
Reviewed-by John FerlanRe: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 15 lines · Yi Min Zhao
Quoted 12 lines · Ján Tomko
Quoted 10 lines · Eduardo Otubo
Quoted 4 lines · Ján Tomko
But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make
seccomp_sandbox= 0 a no-op if it's compiled out.This looks like a good solution for the libvirt side. Can you add this support
so we can merge this fix?
Quoted 5 lines
Quoted 4 lines
Patches proposed:
https://www.redhat.com/archives/libvir-list/2018-May/msg01430.htmlJano
Thanks for your work!
Now pushed in libvirt master:
commit b87222a90919040c12fb6d7c8dcc20f944a66495
Author: Ján Tomko <[email protected]> AuthorDate: 2018-05-18 14:57:51 +0200 Commit: Ján Tomko <[email protected]> CommitDate: 2018-05-23 09:45:48 +0200
qemu: only pass -sandbox off if supported
This way we don't rely on QEMU supplying the -sandbox option without CONFIG_SECCOMP.
git describe: v4.3.0-258-gb87222a909
https://libvirt.org/git/?p=libvirt.git;a=commitdiff;h=b87222a90919040c12fb6d7c8dcc20f944a66495
Jano
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 34 lines · Ján Tomko
Quoted 15 lines · Yi Min Zhao
Quoted 12 lines · Ján Tomko
Quoted 10 lines · Eduardo Otubo
Quoted 4 lines · Ján Tomko
But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make
seccomp_sandbox= 0 a no-op if it's compiled out.This looks like a good solution for the libvirt side. Can you add this support
so we can merge this fix?Quoted 5 lines
Quoted 4 lines
Patches proposed:
https://www.redhat.com/archives/libvir-list/2018-May/msg01430.htmlJano
Thanks for your work!
Now pushed in libvirt master:
commitb87222a90919040c12fb6d7c8dcc20f944a66495Author: Ján Tomko <[email protected]>
AuthorDate: 2018-05-18 14:57:51 +0200 Commit: Ján Tomko <[email protected]>
CommitDate: 2018-05-23 09:45:48 +0200qemu: only pass -sandbox off if supported
This way we don't rely on QEMU supplying the -sandbox option without
CONFIG_SECCOMP.
Quoted 8 lines
Signed-off-by Ján Tomko <[email protected]>Reviewed-by John Ferlan <[email protected]>git describe: v4.3.0-258-gb87222a909
https://libvirt.org/git/?p=libvirt.git;a=commitdiff;h=b87222a90919040c12fb6d7c8dcc20f944a66495Jano
Thanks! But I have not got response from Paolo. I have added him to CC list.
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 40 lines · Yi Min Zhao
Quoted 37 lines · Ján Tomko
Quoted 15 lines · Yi Min Zhao
Quoted 12 lines · Ján Tomko
Quoted 10 lines · Eduardo Otubo
Quoted 4 lines · Ján Tomko
But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make
seccomp_sandbox= 0 a no-op if it's compiled out.This looks like a good solution for the libvirt side. Can you add this support
so we can merge this fix?Quoted 5 lines
Quoted 4 lines
Patches proposed:
https://www.redhat.com/archives/libvir-list/2018-May/msg01430.htmlJano
Thanks for your work!
Now pushed in libvirt master:
commitb87222a90919040c12fb6d7c8dcc20f944a66495Author: Ján Tomko <[email protected]>
AuthorDate: 2018-05-18 14:57:51 +0200 Commit: Ján Tomko <[email protected]>
CommitDate: 2018-05-23 09:45:48 +0200qemu: only pass -sandbox off if supported
This way we don't rely on QEMU supplying the -sandbox option without
CONFIG_SECCOMP.Signed-off-by Ján Tomko <[email protected]>Reviewed-by John Ferlan <[email protected]>
Quoted 7 lines
Quoted 5 lines
git describe: v4.3.0-258-gb87222a909
https://libvirt.org/git/?p=libvirt.git;a=commitdiff;h=b87222a90919040c12fb6d7c8dcc20f944a66495Jano
Thanks! But I have not got response from Paolo. I have added him to CC list.
I'll just wait one more ACK and will send a pull request on the
seccomp queue. Thanks for the contribution.
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 41 lines · Eduardo Otubo
Quoted 40 lines · Yi Min Zhao
Quoted 37 lines · Ján Tomko
Quoted 15 lines · Yi Min Zhao
Quoted 12 lines · Ján Tomko
Quoted 10 lines · Eduardo Otubo
Quoted 4 lines · Ján Tomko
But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make
seccomp_sandbox= 0 a no-op if it's compiled out.This looks like a good solution for the libvirt side. Can you add this support
so we can merge this fix?Quoted 5 lines
Quoted 4 lines
Patches proposed:
https://www.redhat.com/archives/libvir-list/2018-May/msg01430.htmlJano
Thanks for your work!
Now pushed in libvirt master:
commitb87222a90919040c12fb6d7c8dcc20f944a66495Author: Ján Tomko <[email protected]>
AuthorDate: 2018-05-18 14:57:51 +0200 Commit: Ján Tomko <[email protected]>
CommitDate: 2018-05-23 09:45:48 +0200qemu: only pass -sandbox off if supported
This way we don't rely on QEMU supplying the -sandbox option without
CONFIG_SECCOMP.Signed-off-by Ján Tomko <[email protected]>Reviewed-by John Ferlan <[email protected]>
Quoted 10 lines
Quoted 7 lines
Quoted 5 lines
git describe: v4.3.0-258-gb87222a909
https://libvirt.org/git/?p=libvirt.git;a=commitdiff;h=b87222a90919040c12fb6d7c8dcc20f944a66495Jano
Thanks! But I have not got response from Paolo. I have added him to CC list.
I'll just wait one more ACK and will send a pull request on the seccomp queue. Thanks for the contribution.
So... what I should do is wait?
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 48 lines · Yi Min Zhao
Quoted 45 lines · Eduardo Otubo
Quoted 44 lines · Yi Min Zhao
Quoted 41 lines · Ján Tomko
Quoted 15 lines · Yi Min Zhao
Quoted 12 lines · Ján Tomko
Quoted 10 lines · Eduardo Otubo
Quoted 4 lines · Ján Tomko
But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make
seccomp_sandbox= 0 a no-op if it's compiled out.This looks like a good solution for the libvirt side. Can you add this support
so we can merge this fix?Quoted 5 lines
Quoted 4 lines
Patches proposed:
https://www.redhat.com/archives/libvir-list/2018-May/msg01430.htmlJano
Thanks for your work!
Now pushed in libvirt master:
commitb87222a90919040c12fb6d7c8dcc20f944a66495Author: Ján Tomko <[email protected]>
AuthorDate: 2018-05-18 14:57:51 +0200 Commit: Ján Tomko <[email protected]>
CommitDate: 2018-05-23 09:45:48 +0200qemu: only pass -sandbox off if supported
This way we don't rely on QEMU supplying the -sandbox option without
CONFIG_SECCOMP.Signed-off-by Ján Tomko <[email protected]>Reviewed-by John Ferlan <[email protected]>git describe: v4.3.0-258-gb87222a909
https://libvirt.org/git/?p=libvirt.git;a=commitdiff;h=b87222a90919040c12fb6d7c8dcc20f944a66495
Quoted 8 lines
Quoted 6 lines
Quoted 3 lines
Quoted 1 line
Jano
Thanks! But I have not got response from Paolo. I have added him to CC list.
I'll just wait one more ACK and will send a pull request on the seccomp queue. Thanks for the contribution.
So... what I should do is wait?
Yes, even though I think we're safe to proceed without his explicit ack.
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 11 lines · Eduardo Otubo
Quoted 10 lines
Quoted 8 lines
Quoted 4 lines
Quoted 1 line
Thanks! But I have not got response from Paolo. I have added him to CC list.
I'll just wait one more ACK and will send a pull request on the seccomp queue. Thanks for the contribution.
So... what I should do is wait?
Yes, even though I think we're safe to proceed without his explicit ack.
The patch is okay; however, as a follow-up, you could consider moving all the CONFIG_SECCOMP code to qemu-seccomp.c.
This way, the only #ifdef remains the one around qemu_opts_foreach.
Paolo
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 10 lines · Paolo Bonzini
Quoted 9 lines · Eduardo Otubo
Quoted 9 lines
Quoted 7 lines
Quoted 3 lines
Thanks! But I have not got response from Paolo. I have added him to CC list.
I'll just wait one more ACK and will send a pull request on the
seccomp queue. Thanks for the contribution.So... what I should do is wait?
Quoted 9 lines
Quoted 1 line
Yes, even though I think we're safe to proceed without his explicit ack.
The patch is okay; however, as a follow-up, you could consider moving all the
CONFIG_SECCOMPcode toqemu-seccomp.c.This way, the only #ifdef remains the one around
qemu_opts_foreach.Paolo
Thanks for your comment! Indeed, moving to the single C file is much more clear. I will do this after this patch.
@Otubo, what about next step?
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 19 lines · Yi Min Zhao
Quoted 16 lines · Paolo Bonzini
Quoted 10 lines · Eduardo Otubo
Quoted 9 lines
Quoted 7 lines
Quoted 3 lines
Thanks! But I have not got response from Paolo. I have added him to CC list.
I'll just wait one more ACK and will send a pull request on the seccomp queue. Thanks for the contribution.
So... what I should do is wait?
Yes, even though I think we're safe to proceed without his explicit ack.
The patch is okay; however, as a follow-up, you could consider moving all the
CONFIG_SECCOMPcode toqemu-seccomp.c.This way, the only #ifdef remains the one around
qemu_opts_foreach.
Quoted 9 lines
Quoted 3 lines
Paolo
Thanks for your comment! Indeed, moving to the single C file is much more clear. I will do this after this patch.
@Otubo, what about next step?
If you're willing to send v3 with the changes Paolo suggested, I can wait to send the pull request. No worries.
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
Quoted 23 lines · Eduardo Otubo
Quoted 22 lines · Yi Min Zhao
Quoted 19 lines · Paolo Bonzini
Quoted 10 lines · Eduardo Otubo
Quoted 9 lines
Quoted 7 lines
Quoted 3 lines
Thanks! But I have not got response from Paolo. I have added him to CC list.
I'll just wait one more ACK and will send a pull request on the seccomp queue. Thanks for the contribution.
So... what I should do is wait?
Yes, even though I think we're safe to proceed without his explicit ack.
The patch is okay; however, as a follow-up, you could consider moving all the
CONFIG_SECCOMPcode toqemu-seccomp.c.This way, the only #ifdef remains the one around
qemu_opts_foreach.Paolo
Quoted 9 lines
Quoted 6 lines
Thanks for your comment! Indeed, moving to the single C file is much more clear. I will do this after this patch.
@Otubo, what about next step?
If you're willing to send v3 with the changes Paolo suggested, I can wait to send the pull request. No worries.
OK. I will update the new version with Paolo's suggestion.
Re: [Qemu-devel] [PATCH v2 1/1] sandbox: disable -sandbox if CONFIG_SECCOMP undefined
In reply to Ján TomkoQuoted 7 lines · Ján Tomko
This patch fixes the usage of QEMU new enough for seccomp blacklist (where libvirt enables the sandbox by default),
but breaks the usage of QEMU with compiled out sandbox and setting
seccomp_sandbox = 0
in libvirt's qemu.conf:
Quoted 8 lines
error: internal error: process exited while connecting to monitor:
qemu-git: -sandbox off: There is no option group 'sandbox'But now libvirt requires QEMU >= 1.5.0 which already supports query-command-line-options, so if you want the option gone completely --without-seccomp, I can add the code that probes for it and make
seccomp_sandbox= 0 a no-op if it's compiled out.
And that's acceptable - we document that libvirt must be at least as new as qemu. Mixing old qemu + new libvirt should always work, but mixing new qemu + old libvirt may fail, and this is one of those cases. The solution for anyone hitting the failure is to upgrade libvirt to match the fact that they upgraded qemu.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org