xen: Arch Teardown
cover xen: Arch Teardown
Sorry this is late... Patch 1 ought to be ready in this form.
I'll leave it up to the ARM maintainers as to whether they want to take patch 2 in that form (with TEE in patch 3), or whether to merge the two.
Compile tested on all architectures, functionally tested on x86 only.
Branch available from:
https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h=refs/heads/arch-teardown
Full Gitlab CI run in progress:
https://gitlab.com/xen-project/people/andyhhp/xen/-/pipelines/889759838
Andrew Cooper (2):
x86: Introduce arch_domain_teardown()
arm: Boilerpate arch_domain_teardown()
xen/arch/arm/domain.c | 32 ++++++++++++++++++++++++++++++++ xen/arch/x86/domain.c | 5 +++++ xen/common/domain.c | 6 ++++++ xen/include/xen/domain.h | 1 + xen/include/xen/sched.h | 1 + 5 files changed, 45 insertions(+)
base-commit: b3880c365db89051728e1de6b6889c750cbdd915
Signature · 1 line
2.30.2
PATCH 1/2 x86: Introduce arch_domain_teardown()
Plumb it into domain_teardown(). Provide arch_val in the teardown continuation information for use by arch_domain_teardown().
No practical change.
---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Wei Liu <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Jens Wiklander <[email protected]>
--- xen/arch/arm/domain.c | 5 +++++ xen/arch/x86/domain.c | 5 +++++ xen/common/domain.c | 6 ++++++ xen/include/xen/domain.h | 1 + xen/include/xen/sched.h | 1 + 5 files changed, 18 insertions(+)
xen/arch/arm/domain.c +5 -0
@@ -750,6 +750,11 @@ int arch_domain_create(struct domain *d,return rc; } +int arch_domain_teardown (struct domain *d) +{ +return 0 ; +} +void arch_domain_destroy (struct domain *d) {/* IOMMU page table is shared with P2M, always call
xen/arch/x86/domain.c +5 -0
@@ -888,6 +888,11 @@ int arch_domain_create(struct domain *d,return rc; } +int arch_domain_teardown (struct domain *d) +{ +return 0 ; +} +void arch_domain_destroy (struct domain *d) {if (is_hvm_domain (d) )
xen/common/domain.c +6 -0
@@ -416,6 +416,7 @@ static int domain_teardown(struct domain *d) PROG_none, PROG_gnttab_mappings, PROG_vcpu_teardown, + PROG_arch_teardown, PROG_done, };
@@ -436,6 +437,11 @@ static int domain_teardown(struct domain *d)return rc; } + PROGRESS(arch_teardown): + rc =arch_domain_teardown (d); +if ( rc ) +return rc; + PROGRESS(done): break;
xen/include/xen/domain.h +1 -0
@@ -80,6 +80,7 @@ int arch_domain_create(struct domain *d,struct xen_domctl_createdomain *config, unsigned int flags); +int arch_domain_teardown (struct domain *d);void arch_domain_destroy (struct domain *d);void arch_domain_shutdown (struct domain *d);
xen/include/xen/sched.h +1 -0
@@ -589,6 +589,7 @@ struct domain*/ struct {unsigned int val; +unsigned int arch_val;struct vcpu *vcpu; } teardown;
Signature · 1 line
2.30.2
Re: [PATCH 1/2] x86: Introduce arch_domain_teardown()
Quoted 96 lines · Andrew Cooper
Plumb it into
domain_teardown(). Providearch_valin the teardown continuation information for use byarch_domain_teardown().No practical change.
Signed-off-by Andrew Cooper <[email protected]>---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Wei Liu <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Jens Wiklander <[email protected]>--- xen/arch/arm/domain.c | 5 +++++ xen/arch/x86/domain.c | 5 +++++ xen/common/domain.c | 6 ++++++ xen/include/xen/domain.h | 1 + xen/include/xen/sched.h | 1 + 5 files changed, 18 insertions(+)xen/arch/arm/domain.c +5
-0@@ -750,6 +750,11 @@ int arch_domain_create(struct domain *d,return rc; } +int arch_domain_teardown (struct domain *d) +{ +return 0 ; +} +void arch_domain_destroy (struct domain *d) {/* IOMMU page table is shared with P2M, always call xen/arch/x86/domain.c +5
-0@@ -888,6 +888,11 @@ int arch_domain_create(struct domain *d,return rc; } +int arch_domain_teardown (struct domain *d) +{ +return 0 ; +} +void arch_domain_destroy (struct domain *d) {if (is_hvm_domain (d) )xen/common/domain.c +6
-0@@ -416,6 +416,7 @@ static int domain_teardown(struct domain *d) PROG_none, PROG_gnttab_mappings, PROG_vcpu_teardown, + PROG_arch_teardown, PROG_done, };@@ -436,6 +437,11 @@ static int domain_teardown(struct domain *d)return rc; } + PROGRESS(arch_teardown): + rc =arch_domain_teardown (d); +if ( rc ) +return rc; + PROGRESS(done): break;xen/include/xen/domain.h +1
-0@@ -80,6 +80,7 @@ int arch_domain_create(struct domain *d,struct xen_domctl_createdomain *config, unsigned int flags); +int arch_domain_teardown (struct domain *d);void arch_domain_destroy (struct domain *d);void arch_domain_shutdown (struct domain *d);xen/include/xen/sched.h +0
-0
xen/include/xen/sched.h struct domain · line 592
Quoted 4 lines
xen/include/xen/sched.h +1
-0@@ -589,6 +589,7 @@ struct domain*/ struct {unsigned int val; +unsigned int arch_val;
While I haven't looked at patch 2, wouldn't such continuation information better be encoded in arch_domain in whatever way is more suitable for each architecture?
Thanks, Roger.
Re: [PATCH 1/2] x86: Introduce arch_domain_teardown()
Quoted 13 lines · Roger Pau Monné
xen/include/xen/sched.h struct domain · line 592
Quoted 9 lines · Andrew Cooper
xen/include/xen/sched.h +1
-0@@ -589,6 +589,7 @@ struct domain*/ struct {unsigned int val; +unsigned int arch_val;While I haven't looked at patch 2, wouldn't such continuation information better be encoded in
arch_domainin whatever way is more suitable for each architecture?
Well, it's filling a hole here on 64bit builds, which it couldn't do in arch_domain.
And it's contained inside teardown{} which already signals it as fairly magic.
~Andrew
Jan Beulich
Acked-byRe: [PATCH 1/2] x86: Introduce arch_domain_teardown()
Quoted 15 lines · Andrew Cooper
Quoted 13 lines · Roger Pau Monné
xen/include/xen/sched.h struct domain · line 592
Quoted 9 lines · Andrew Cooper
xen/include/xen/sched.h +1
-0@@ -589,6 +589,7 @@ struct domain*/ struct {unsigned int val; +unsigned int arch_val;While I haven't looked at patch 2, wouldn't such continuation information better be encoded in
arch_domainin whatever way is more suitable for each architecture?
Quoted 5 lines
Well, it's filling a hole here on 64bit builds, which it couldn't do in
arch_domain.And it's contained inside teardown{} which already signals it as fairly magic.
Plus why have each architecture duplicate the field? I expect none of the arch_domain_teardown() instances will remain without an actual use of the new field, mid to long term.
I don't want to override Roger's concern, but from my pov
Jan
Re: [PATCH 1/2] x86: Introduce arch_domain_teardown()
Quoted 22 lines · Jan Beulich
Quoted 15 lines · Andrew Cooper
Quoted 13 lines · Roger Pau Monné
xen/include/xen/sched.h struct domain · line 592
Quoted 9 lines · Andrew Cooper
xen/include/xen/sched.h +1
-0@@ -589,6 +589,7 @@ struct domain*/ struct {unsigned int val; +unsigned int arch_val;While I haven't looked at patch 2, wouldn't such continuation information better be encoded in
arch_domainin whatever way is more suitable for each architecture?Quoted 5 lines
Well, it's filling a hole here on 64bit builds, which it couldn't do in
arch_domain.And it's contained inside teardown{} which already signals it as fairly magic.
Quoted 6 lines
Plus why have each architecture duplicate the field? I expect none of the
arch_domain_teardown()instances will remain without an actual use of the new field, mid to long term.I don't want to override Roger's concern, but from my pov
Acked-by Jan Beulich <[email protected]>
Oh, no worries, I was meaning to reply I was fine with Andrews justification, but forgot.
Thanks, Roger
PATCH 2/2 arm: Boilerpate arch_domain_teardown()
In reply to Andrew CooperXXX to be filled in with TEE teardown.
---
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bertrand Marquis <[email protected]>
CC: Jens Wiklander <[email protected]>
Jens: In the same was as the previous patch in the common path, you want to add a PROG_tee(?) here, and rearrange the right function(s).
--- xen/arch/arm/domain.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)
xen/arch/arm/domain.c +27 -0
@@ -752,6 +752,33 @@ int arch_domain_create(struct domain *d,int arch_domain_teardown (struct domain *d) { +BUG_ON (!d->is_dying); + +/* See domain_teardown() for an explanation of all of this magic. */ +switch ( d->teardown.arch_val ) + { +#define PROGRESS (x) \ + d->teardown.arch_val = PROG_ ## x; \ + fallthrough; \ + case PROG_ ## x + +enum { + PROG_none, + PROG_done, + }; + +case PROG_none: +BUILD_BUG_ON (PROG_none !=0 ); + + PROGRESS(done): + break; + +#undef PROGRESS + +default : +BUG (); + } +return 0 ; }
Signature · 1 line
2.30.2
Hi Andrew,
Quoted 9 lines · Andrew Cooper
Sorry this is late... Patch 1 ought to be ready in this form.
I'll leave it up to the ARM maintainers as to whether they want to take patch 2 in that form (with TEE in patch 3), or whether to merge the two.
Compile tested on all architectures, functionally tested on x86 only.
Quoted 7 lines
Branch available from:
https://xenbits.xen.org/gitweb/?p=people/andrewcoop/xen.git;a=shortlog;h=refs/heads/arch-teardown
Full Gitlab CI run in progress:
https://gitlab.com/xen-project/people/andyhhp/xen/-/pipelines/889759838
Thanks a lot for that.
I will sync up with Jens and we will handle the rest :-)
Cheers
Bertrand
Quoted 17 lines
Andrew Cooper (2):
x86: Introducearch_domain_teardown()
arm: Boilerpatearch_domain_teardown()xen/arch/arm/domain.c | 32 ++++++++++++++++++++++++++++++++ xen/arch/x86/domain.c | 5 +++++ xen/common/domain.c | 6 ++++++ xen/include/xen/domain.h | 1 + xen/include/xen/sched.h | 1 + 5 files changed, 45 insertions(+)base-commit:
b3880c365db89051728e1de6b6889c750cbdd915Signature · 1 line
2.30.2