hyperv: root partition faults writing to VP ASSIST MSR PAGE
Subsystems (3)
PATCH v3 1/1 hyperv: root partition faults writing to VP ASSIST MSR PAGE
For Root partition the VP assist pages are pre-determined by the hypervisor. The Root kernel is not allowed to change them to different locations. And thus, we are getting below stack as in current implementation Root is trying to perform write to specific MSR.
[ 2.778197] unchecked MSR access error: WRMSR to 0x40000073 (tried to write 0x0000000145ac5001) at rIP: 0xffffffff810c1084 (native_write_msr+0x4/0x30) [ 2.784867] Call Trace: [ 2.791507] hv_cpu_init+0xf1/0x1c0 [ 2.798144] ? hyperv_report_panic+0xd0/0xd0 [ 2.804806] cpuhp_invoke_callback+0x11a/0x440 [ 2.811465] ? hv_resume+0x90/0x90 [ 2.818137] cpuhp_issue_call+0x126/0x130 [ 2.824782] __cpuhp_setup_state_cpuslocked+0x102/0x2b0 [ 2.831427] ? hyperv_report_panic+0xd0/0xd0 [ 2.838075] ? hyperv_report_panic+0xd0/0xd0 [ 2.844723] ? hv_resume+0x90/0x90 [ 2.851375] __cpuhp_setup_state+0x3d/0x90 [ 2.858030] hyperv_init+0x14e/0x410 [ 2.864689] ? enable_IR_x2apic+0x190/0x1a0 [ 2.871349] apic_intr_mode_init+0x8b/0x100 [ 2.878017] x86_late_time_init+0x20/0x30 [ 2.884675] start_kernel+0x459/0x4fb [ 2.891329] secondary_startup_64_no_verify+0xb0/0xbb
Since, the hypervisor already provides the VP assist page for root partition, we need to memremap the memory from hypervisor for root kernel to use. The mapping is done in hv_cpu_init during bringup and is unmaped in hv_cpu_die during teardown.
--- arch/x86/hyperv/hv_init.c | 61 +++++++++++++++++++++--------- arch/x86/include/asm/hyperv-tlfs.h | 9 +++++ 2 files changed, 53 insertions(+), 17 deletions(-)
changelog:
v1: initial patch
v2: commit message changes, removal of HV_MSR_APIC_ACCESS_AVAILABLE
check and addition of null check before reading the VP assist MSR for root partition
v3: added new data structure to handle VP ASSIST MSR page and done
handling in hv_cpu_init and hv_cpu_die
---
arch/x86/hyperv/hv_init.c +44 -17
@@ -44,6 +44,7 @@ EXPORT_SYMBOL_GPL(hv_vp_assist_page);static int hv_cpu_init (unsigned int cpu) { +union hv_vp_assist_msr_contents msr;struct hv_vp_assist_page **hvp = &hv_vp_assist_page[smp_processor_id ()];int ret;
@@ -54,27 +55,41 @@ static int hv_cpu_init(unsigned int cpu)if (!hv_vp_assist_page)return 0 ; -/* -* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's Section -* 5.2.1 "GPA Overlay Pages"). Here it must be zeroed out to make sure -* we always write the EOI MSR in hv_apic_eoi_write() *after* the -* EOI optimization is disabled in hv_cpu_die(), otherwise a CPU may -* not be stopped in the case of CPU offlining and the VM will hang. -*/ -if (!*hvp) { - *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); +if (hv_root_partition) { +/* +* For Root partition we get the hypervisor provided VP ASSIST +* PAGE, instead of allocating a new page. +*/ +rdmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64); + +/* remapping to root partition address space */ +if (!*hvp) + *hvp =memremap (msr.guest_physical_address << +HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT , +PAGE_SIZE ,MEMREMAP_WB ); + }else { +/* +* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's +* Section 5.2.1 "GPA Overlay Pages"). Here it must be zeroed +* out to make sure we always write the EOI MSR in +* hv_apic_eoi_write() *after* theEOI optimization is disabled +* in hv_cpu_die(), otherwise a CPU may not be stopped in the +* case of CPU offlining and the VM will hang. +*/ +if (!*hvp) + *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); + } -if (*hvp) { -u64 val; +WARN_ON (!(*hvp)); - val =vmalloc_to_pfn (*hvp); - val = (val <<HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT ) | -HV_X64_MSR_VP_ASSIST_PAGE_ENABLE ; +if (*hvp) { +if (!hv_root_partition) + msr.guest_physical_address =vmalloc_to_pfn (*hvp); -wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , val); + msr.enable =1 ; +wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64); } -return 0 ; }
@@ -170,9 +185,21 @@ static int hv_cpu_die(unsigned int cpu)hv_common_cpu_die (cpu); -if (hv_vp_assist_page && hv_vp_assist_page[cpu]) +if (hv_vp_assist_page && hv_vp_assist_page[cpu]) {wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE ,0 ); +if (hv_root_partition) { +/* +* For Root partition the VP ASSIST page is mapped to +* hypervisor provided page, and thus, we unmap the +* page here and nullify it, so that in future we have +* correct page address mapped in hv_cpu_init +*/ +memunmap (hv_vp_assist_page[cpu]); + hv_vp_assist_page[cpu] =NULL ; + } + } +if (hv_reenlightenment_cb ==NULL )return 0 ;
arch/x86/include/asm/hyperv-tlfs.h +9 -0
@@ -288,6 +288,15 @@ union hv_x64_msr_hypercall_contents { } __packed; }; +union hv_vp_assist_msr_contents { +u64 as_uint64; +struct { +u64 enable:1 ; +u64 reserved:11 ; +u64 guest_physical_address:52 ; + } __packed; +}; +struct hv_reenlightenment_control {__u64 vector:8 ;__u64 reserved1:8 ;
Signature · 1 line
2.25.1
From: Praveen Kumar <[email protected]> Sent: Tuesday, July 27, 2021 3:41 AM
Quoted 88 lines
For Root partition the VP assist pages are pre-determined by the hypervisor. The Root kernel is not allowed to change them to different locations. And thus, we are getting below stack as in current implementation Root is trying to perform write to specific MSR.
[ 2.778197] unchecked MSR access error: WRMSR to 0x40000073 (tried to write 0x0000000145ac5001) at rIP: 0xffffffff810c1084 (native_write_msr+0x4/0x30) [ 2.784867] Call Trace: [ 2.791507] hv_cpu_init+0xf1/0x1c0 [ 2.798144] ? hyperv_report_panic+0xd0/0xd0 [ 2.804806] cpuhp_invoke_callback+0x11a/0x440 [ 2.811465] ? hv_resume+0x90/0x90 [ 2.818137] cpuhp_issue_call+0x126/0x130 [ 2.824782] __cpuhp_setup_state_cpuslocked+0x102/0x2b0 [ 2.831427] ? hyperv_report_panic+0xd0/0xd0 [ 2.838075] ? hyperv_report_panic+0xd0/0xd0 [ 2.844723] ? hv_resume+0x90/0x90 [ 2.851375] __cpuhp_setup_state+0x3d/0x90 [ 2.858030] hyperv_init+0x14e/0x410 [ 2.864689] ? enable_IR_x2apic+0x190/0x1a0 [ 2.871349] apic_intr_mode_init+0x8b/0x100 [ 2.878017] x86_late_time_init+0x20/0x30 [ 2.884675] start_kernel+0x459/0x4fb [ 2.891329] secondary_startup_64_no_verify+0xb0/0xbbSince, the hypervisor already provides the VP assist page for root partition, we need to memremap the memory from hypervisor for root kernel to use. The mapping is done in
hv_cpu_initduring bringup and is unmaped inhv_cpu_dieduring teardown.Signed-off-by Praveen Kumar <[email protected]>--- arch/x86/hyperv/hv_init.c | 61 +++++++++++++++++++++--------- arch/x86/include/asm/hyperv-tlfs.h | 9 +++++ 2 files changed, 53 insertions(+), 17 deletions(-)changelog:
v1: initial patch
v2: commit message changes, removal ofHV_MSR_APIC_ACCESS_AVAILABLE
check and addition of null check before reading the VP assist MSR for root partition
v3: added new data structure to handle VP ASSIST MSR page and done
handling inhv_cpu_initandhv_cpu_die---
arch/x86/hyperv/hv_init.c +16
-9@@ -44,6 +44,7 @@ EXPORT_SYMBOL_GPL(hv_vp_assist_page);static int hv_cpu_init (unsigned int cpu) { +union hv_vp_assist_msr_contents msr;struct hv_vp_assist_page **hvp = &hv_vp_assist_page[smp_processor_id ()];int ret;@@ -54,27 +55,41 @@ static int hv_cpu_init(unsigned int cpu)if (!hv_vp_assist_page)return 0 ; -/* -* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's Section -* 5.2.1 "GPA Overlay Pages"). Here it must be zeroed out to make sure -* we always write the EOI MSR in hv_apic_eoi_write() *after* the -* EOI optimization is disabled in hv_cpu_die(), otherwise a CPU may -* not be stopped in the case of CPU offlining and the VM will hang. -*/ -if (!*hvp) { - *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); +if (hv_root_partition) { +/* +* For Root partition we get the hypervisor provided VP ASSIST +* PAGE, instead of allocating a new page. +*/ +rdmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64); + +/* remapping to root partition address space */ +if (!*hvp) + *hvp =memremap (msr.guest_physical_address << +HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT , +PAGE_SIZE ,MEMREMAP_WB ); + }else { +/* +* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 82
Quoted 10 lines
arch/x86/hyperv/hv_init.c +9
-0+* Section 5.2.1 "GPA Overlay Pages"). Here it must be zeroed +* out to make sure we always write the EOI MSR in +* hv_apic_eoi_write() *after* theEOI optimization is disabled +* in hv_cpu_die(), otherwise a CPU may not be stopped in the +* case of CPU offlining and the VM will hang. +*/ +if (!*hvp) + *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); + }
The tests here could be reversed to eliminate some duplication. For example:
if(!*hvp) {
if (hv_root_partition) {
rdmsrl(....);
*hvp = memremap( .....);
} else {
*hvp = __vmalloc(....);
}
}Quoted 5 lines
arch/x86/hyperv/hv_init.c +1
-2-if (*hvp) { -u64 val; +WARN_ON (!(*hvp));
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 91
Quoted 10 lines
arch/x86/hyperv/hv_init.c +5
-4- val =vmalloc_to_pfn (*hvp); - val = (val <<HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT ) | -HV_X64_MSR_VP_ASSIST_PAGE_ENABLE ; +if (*hvp) { +if (!hv_root_partition) + msr.guest_physical_address =vmalloc_to_pfn (*hvp); -wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , val); + msr.enable =1 ; +wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64);
This version has a substantive difference compared with previous versions in that the "enable" bit is being set and written back to the MSR even when running in the root partition. Is that intentional?
arch/x86/hyperv/hv_init.c static int hv_cpu_die(unsigned int cpu) · line 189
Quoted 12 lines
arch/x86/hyperv/hv_init.c +1
-2} -return 0 ; }@@ -170,9 +185,21 @@ static int hv_cpu_die(unsigned int cpu)hv_common_cpu_die (cpu); -if (hv_vp_assist_page && hv_vp_assist_page[cpu]) +if (hv_vp_assist_page && hv_vp_assist_page[cpu]) {wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE ,0 );
This will set the guest_physical_address in the MSR to zero, even in the root partition case. Is that OK? It seems inconsistent with hv_cpu_init() where the existing guest_physical_address in the MSR is carefully preserved for the root partition case. Or is the intent here simply to clear the "enable" flag?
Quoted 21 lines
arch/x86/hyperv/hv_init.c +12
-0+if (hv_root_partition) { +/* +* For Root partition the VP ASSIST page is mapped to +* hypervisor provided page, and thus, we unmap the +* page here and nullify it, so that in future we have +* correct page address mapped in hv_cpu_init +*/ +memunmap (hv_vp_assist_page[cpu]); + hv_vp_assist_page[cpu] =NULL ; + } + } +if (hv_reenlightenment_cb ==NULL )return 0 ;arch/x86/include/asm/hyperv-tlfs.h +0
-0
arch/x86/include/asm/hyperv-tlfs.h union hv_x64_msr_hypercall_contents { · line 296
Quoted 9 lines
arch/x86/include/asm/hyperv-tlfs.h +6
-0@@ -288,6 +288,15 @@ union hv_x64_msr_hypercall_contents { } __packed; }; +union hv_vp_assist_msr_contents { +u64 as_uint64; +struct { +u64 enable:1 ; +u64 reserved:11 ; +u64 guest_physical_address:52 ;
This field really should be named "guest_physical_page", as it is a page number, not an address. You've matched the field names used in hv_x64_msr_hypercall_contents, which is good for consistency, except that the field name is
wrong in hv_x64_msr_hypercall_contents. :-( I think
the Hyper-V TLFS originally called it a "physical address", but the TLFS has since been fixed to described it as a page number. I'd suggest getting this one named correctly; fixing the field name in hv_x64_msr_hypercall_contents is a separate cleanup that doesn't need to be done now.
Quoted 8 lines
arch/x86/include/asm/hyperv-tlfs.h +3
-1+ } __packed; +}; +struct hv_reenlightenment_control {__u64 vector:8 ;__u64 reserved1:8 ; --2.25.1
Quoted 12 lines
Quoted 12 lines
arch/x86/hyperv/hv_init.c +4
-5-if (*hvp) { -u64 val; +WARN_ON (!(*hvp)); - val =vmalloc_to_pfn (*hvp); - val = (val <<HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT ) | -HV_X64_MSR_VP_ASSIST_PAGE_ENABLE ; +if (*hvp) { +if (!hv_root_partition) + msr.guest_physical_address =vmalloc_to_pfn (*hvp);
Quoted 7 lines
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 91
Quoted 3 lines
arch/x86/hyperv/hv_init.c +2
-1-wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , val); + msr.enable =1 ; +wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64);This version has a substantive difference compared with previous versions in that the "enable" bit is being set and written back to the MSR even when running in the root partition. Is that intentional?
Yes, the hypervisor won't allow the root to change the page address. So, to enable/disable it, the page address needs to be preserved and just the enable bit needs to be toggled.
- Sunil
Quoted 102 lines · Michael Kelley
From: Praveen Kumar <[email protected]> Sent: Tuesday, July 27, 2021 3:41 AM
Quoted 88 lines
For Root partition the VP assist pages are pre-determined by the hypervisor. The Root kernel is not allowed to change them to different locations. And thus, we are getting below stack as in current implementation Root is trying to perform write to specific MSR.
[ 2.778197] unchecked MSR access error: WRMSR to 0x40000073 (tried to write 0x0000000145ac5001) at rIP: 0xffffffff810c1084 (native_write_msr+0x4/0x30) [ 2.784867] Call Trace: [ 2.791507] hv_cpu_init+0xf1/0x1c0 [ 2.798144] ? hyperv_report_panic+0xd0/0xd0 [ 2.804806] cpuhp_invoke_callback+0x11a/0x440 [ 2.811465] ? hv_resume+0x90/0x90 [ 2.818137] cpuhp_issue_call+0x126/0x130 [ 2.824782] __cpuhp_setup_state_cpuslocked+0x102/0x2b0 [ 2.831427] ? hyperv_report_panic+0xd0/0xd0 [ 2.838075] ? hyperv_report_panic+0xd0/0xd0 [ 2.844723] ? hv_resume+0x90/0x90 [ 2.851375] __cpuhp_setup_state+0x3d/0x90 [ 2.858030] hyperv_init+0x14e/0x410 [ 2.864689] ? enable_IR_x2apic+0x190/0x1a0 [ 2.871349] apic_intr_mode_init+0x8b/0x100 [ 2.878017] x86_late_time_init+0x20/0x30 [ 2.884675] start_kernel+0x459/0x4fb [ 2.891329] secondary_startup_64_no_verify+0xb0/0xbbSince, the hypervisor already provides the VP assist page for root partition, we need to memremap the memory from hypervisor for root kernel to use. The mapping is done in
hv_cpu_initduring bringup and is unmaped inhv_cpu_dieduring teardown.Signed-off-by Praveen Kumar <[email protected]>--- arch/x86/hyperv/hv_init.c | 61 +++++++++++++++++++++--------- arch/x86/include/asm/hyperv-tlfs.h | 9 +++++ 2 files changed, 53 insertions(+), 17 deletions(-)changelog:
v1: initial patch
v2: commit message changes, removal ofHV_MSR_APIC_ACCESS_AVAILABLE
check and addition of null check before reading the VP assist MSR for root partition
v3: added new data structure to handle VP ASSIST MSR page and done
handling inhv_cpu_initandhv_cpu_die---
arch/x86/hyperv/hv_init.c +16
-9@@ -44,6 +44,7 @@ EXPORT_SYMBOL_GPL(hv_vp_assist_page);static int hv_cpu_init (unsigned int cpu) { +union hv_vp_assist_msr_contents msr;struct hv_vp_assist_page **hvp = &hv_vp_assist_page[smp_processor_id ()];int ret;@@ -54,27 +55,41 @@ static int hv_cpu_init(unsigned int cpu)if (!hv_vp_assist_page)return 0 ; -/* -* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's Section -* 5.2.1 "GPA Overlay Pages"). Here it must be zeroed out to make sure -* we always write the EOI MSR in hv_apic_eoi_write() *after* the -* EOI optimization is disabled in hv_cpu_die(), otherwise a CPU may -* not be stopped in the case of CPU offlining and the VM will hang. -*/ -if (!*hvp) { - *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); +if (hv_root_partition) { +/* +* For Root partition we get the hypervisor provided VP ASSIST +* PAGE, instead of allocating a new page. +*/ +rdmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64); + +/* remapping to root partition address space */ +if (!*hvp) + *hvp =memremap (msr.guest_physical_address << +HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT , +PAGE_SIZE ,MEMREMAP_WB ); + }else { +/* +* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 82
Quoted 10 lines
arch/x86/hyperv/hv_init.c +9
-0+* Section 5.2.1 "GPA Overlay Pages"). Here it must be zeroed +* out to make sure we always write the EOI MSR in +* hv_apic_eoi_write() *after* theEOI optimization is disabled +* in hv_cpu_die(), otherwise a CPU may not be stopped in the +* case of CPU offlining and the VM will hang. +*/ +if (!*hvp) + *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); + }The tests here could be reversed to eliminate some duplication. For example:
Quoted 10 lines
if(!*hvp) {
if (hv_root_partition) {
rdmsrl(....);
*hvp = memremap( .....);
} else {
*hvp = __vmalloc(....);
}
}Sure. Thanks.
Quoted 12 lines
Quoted 12 lines
arch/x86/hyperv/hv_init.c +4
-5-if (*hvp) { -u64 val; +WARN_ON (!(*hvp)); - val =vmalloc_to_pfn (*hvp); - val = (val <<HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT ) | -HV_X64_MSR_VP_ASSIST_PAGE_ENABLE ; +if (*hvp) { +if (!hv_root_partition) + msr.guest_physical_address =vmalloc_to_pfn (*hvp);
Quoted 8 lines
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 91
Quoted 3 lines
arch/x86/hyperv/hv_init.c +2
-1-wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , val); + msr.enable =1 ; +wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64);This version has a substantive difference compared with previous versions in that the "enable" bit is being set and written back to the MSR even when running in the root partition. Is that intentional?
Yes, we need to enable the same for root partition as well.
Quoted 52 lines
arch/x86/hyperv/hv_init.c static int hv_cpu_die(unsigned int cpu) · line 189
Quoted 12 lines
arch/x86/hyperv/hv_init.c +1
-2} -return 0 ; }@@ -170,9 +185,21 @@ static int hv_cpu_die(unsigned int cpu)hv_common_cpu_die (cpu); -if (hv_vp_assist_page && hv_vp_assist_page[cpu]) +if (hv_vp_assist_page && hv_vp_assist_page[cpu]) {wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE ,0 );This will set the
guest_physical_addressin the MSR to zero, even in the root partition case. Is that OK? It seems inconsistent withhv_cpu_init()where the existingguest_physical_addressin the MSR is carefully preserved for the root partition case. Or is the intent here simply to clear the "enable" flag?Quoted 21 lines
arch/x86/hyperv/hv_init.c +12
-0+if (hv_root_partition) { +/* +* For Root partition the VP ASSIST page is mapped to +* hypervisor provided page, and thus, we unmap the +* page here and nullify it, so that in future we have +* correct page address mapped in hv_cpu_init +*/ +memunmap (hv_vp_assist_page[cpu]); + hv_vp_assist_page[cpu] =NULL ; + } + } +if (hv_reenlightenment_cb ==NULL )return 0 ;arch/x86/include/asm/hyperv-tlfs.h +0
-0arch/x86/include/asm/hyperv-tlfs.h union hv_x64_msr_hypercall_contents { · line 296
Quoted 9 lines
arch/x86/include/asm/hyperv-tlfs.h +6
-0@@ -288,6 +288,15 @@ union hv_x64_msr_hypercall_contents { } __packed; }; +union hv_vp_assist_msr_contents { +u64 as_uint64; +struct { +u64 enable:1 ; +u64 reserved:11 ; +u64 guest_physical_address:52 ;This field really should be named "
guest_physical_page", as it is a page number, not an address. You've matched the
Quoted 9 lines
field names used in
hv_x64_msr_hypercall_contents, which is good for consistency, except that the field name is
wrong inhv_x64_msr_hypercall_contents. :-( I think
the Hyper-V TLFS originally called it a "physical address", but the TLFS has since been fixed to described it as a page number. I'd suggest getting this one named correctly; fixing the field name inhv_x64_msr_hypercall_contentsis a separate cleanup that doesn't need to be done now.
Sure. Will do it for this new data structure.
Quoted 8 lines
Quoted 8 lines
arch/x86/include/asm/hyperv-tlfs.h +3
-1+ } __packed; +}; +struct hv_reenlightenment_control {__u64 vector:8 ;__u64 reserved1:8 ; --2.25.1
Regards,
~Praveen.
From: Praveen Kumar <[email protected]> Sent: Tuesday, July 27, 2021 3:41 AM
Quoted 61 lines
For Root partition the VP assist pages are pre-determined by the hypervisor. The Root kernel is not allowed to change them to different locations. And thus, we are getting below stack as in current implementation Root is trying to perform write to specific MSR.
[ 2.778197] unchecked MSR access error: WRMSR to 0x40000073 (tried to write 0x0000000145ac5001) at rIP: 0xffffffff810c1084 (native_write_msr+0x4/0x30) [ 2.784867] Call Trace: [ 2.791507] hv_cpu_init+0xf1/0x1c0 [ 2.798144] ? hyperv_report_panic+0xd0/0xd0 [ 2.804806] cpuhp_invoke_callback+0x11a/0x440 [ 2.811465] ? hv_resume+0x90/0x90 [ 2.818137] cpuhp_issue_call+0x126/0x130 [ 2.824782] __cpuhp_setup_state_cpuslocked+0x102/0x2b0 [ 2.831427] ? hyperv_report_panic+0xd0/0xd0 [ 2.838075] ? hyperv_report_panic+0xd0/0xd0 [ 2.844723] ? hv_resume+0x90/0x90 [ 2.851375] __cpuhp_setup_state+0x3d/0x90 [ 2.858030] hyperv_init+0x14e/0x410 [ 2.864689] ? enable_IR_x2apic+0x190/0x1a0 [ 2.871349] apic_intr_mode_init+0x8b/0x100 [ 2.878017] x86_late_time_init+0x20/0x30 [ 2.884675] start_kernel+0x459/0x4fb [ 2.891329] secondary_startup_64_no_verify+0xb0/0xbbSince, the hypervisor already provides the VP assist page for root partition, we need to memremap the memory from hypervisor for root kernel to use. The mapping is done in
hv_cpu_initduring bringup and is unmaped inhv_cpu_dieduring teardown.Signed-off-by Praveen Kumar <[email protected]>--- arch/x86/hyperv/hv_init.c | 61 +++++++++++++++++++++--------- arch/x86/include/asm/hyperv-tlfs.h | 9 +++++ 2 files changed, 53 insertions(+), 17 deletions(-)changelog:
v1: initial patch
v2: commit message changes, removal ofHV_MSR_APIC_ACCESS_AVAILABLE
check and addition of null check before reading the VP assist MSR for root partition
v3: added new data structure to handle VP ASSIST MSR page and done
handling inhv_cpu_initandhv_cpu_die---
arch/x86/hyperv/hv_init.c +1
-0@@ -44,6 +44,7 @@ EXPORT_SYMBOL_GPL(hv_vp_assist_page);static int hv_cpu_init (unsigned int cpu) { +union hv_vp_assist_msr_contents msr;struct hv_vp_assist_page **hvp = &hv_vp_assist_page[smp_processor_id ()];int ret;
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 56
Quoted 2 lines
arch/x86/hyperv/hv_init.c +0
-0@@ -54,27 +55,41 @@ static int hv_cpu_init(unsigned int cpu)if (!hv_vp_assist_page)return 0 ;
Not related to this code, but I am not sure about the usefulness of this NULL check as we have already accessed this pointer above. If it was NULL, things would already blow up.
Quoted 8 lines
arch/x86/hyperv/hv_init.c +0
-7-/* -* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's Section -* 5.2.1 "GPA Overlay Pages"). Here it must be zeroed out to make sure -* we always write the EOI MSR in hv_apic_eoi_write() *after* the -* EOI optimization is disabled in hv_cpu_die(), otherwise a CPU may -* not be stopped in the case of CPU offlining and the VM will hang. -*/
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 65
Quoted 10 lines
arch/x86/hyperv/hv_init.c +8
-2-if (!*hvp) { - *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); +if (hv_root_partition) { +/* +* For Root partition we get the hypervisor provided VP ASSIST +* PAGE, instead of allocating a new page. +*/ +rdmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64); + +/* remapping to root partition address space */
Better to leave out comments that are obvious from the code.
Quoted 18 lines
arch/x86/hyperv/hv_init.c +16
-0+if (!*hvp) + *hvp =memremap (msr.guest_physical_address << +HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT , +PAGE_SIZE ,MEMREMAP_WB ); + }else { +/* +* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's +* Section 5.2.1 "GPA Overlay Pages"). Here it must be zeroed +* out to make sure we always write the EOI MSR in +* hv_apic_eoi_write() *after* theEOI optimization is disabled +* in hv_cpu_die(), otherwise a CPU may not be stopped in the +* case of CPU offlining and the VM will hang. +*/ +if (!*hvp) + *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); + }
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 88
Quoted 10 lines
arch/x86/hyperv/hv_init.c +4
-5-if (*hvp) { -u64 val; +WARN_ON (!(*hvp)); - val =vmalloc_to_pfn (*hvp); - val = (val <<HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT ) | -HV_X64_MSR_VP_ASSIST_PAGE_ENABLE ; +if (*hvp) { +if (!hv_root_partition) + msr.guest_physical_address =vmalloc_to_pfn (*hvp);
It's better to move this above in the else section where we are 'vmalloc' the page. If you just check for the NULL for the page above and return if NULL, that should clean up the code as well.
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 90
Quoted 3 lines
arch/x86/hyperv/hv_init.c +1
-1-wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , val); + msr.enable =1 ;
We should also set the reserved bits to 0.
arch/x86/hyperv/hv_init.c static int hv_cpu_die(unsigned int cpu) · line 189
Quoted 13 lines
arch/x86/hyperv/hv_init.c +2
-2+wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64); } -return 0 ; }@@ -170,9 +185,21 @@ static int hv_cpu_die(unsigned int cpu)hv_common_cpu_die (cpu); -if (hv_vp_assist_page && hv_vp_assist_page[cpu]) +if (hv_vp_assist_page && hv_vp_assist_page[cpu]) {wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE ,0 );
Its better to read the MSR, set the enable bit to 0 and write it back.
arch/x86/hyperv/hv_init.c static int hv_cpu_die(unsigned int cpu) · line 200
Quoted 11 lines
arch/x86/hyperv/hv_init.c +10
-0+if (hv_root_partition) { +/* +* For Root partition the VP ASSIST page is mapped to +* hypervisor provided page, and thus, we unmap the +* page here and nullify it, so that in future we have +* correct page address mapped in hv_cpu_init +*/ +memunmap (hv_vp_assist_page[cpu]); + hv_vp_assist_page[cpu] =NULL ; + }
For the guest case, where are we freeing the page?
Quoted 10 lines
arch/x86/hyperv/hv_init.c +2
-0+ } +if (hv_reenlightenment_cb ==NULL )return 0 ;arch/x86/include/asm/hyperv-tlfs.h +0
-0
arch/x86/include/asm/hyperv-tlfs.h union hv_x64_msr_hypercall_contents { · line 296
Quoted 9 lines
arch/x86/include/asm/hyperv-tlfs.h +6
-0@@ -288,6 +288,15 @@ union hv_x64_msr_hypercall_contents { } __packed; }; +union hv_vp_assist_msr_contents { +u64 as_uint64; +struct { +u64 enable:1 ; +u64 reserved:11 ; +u64 guest_physical_address:52 ;
This field contains the page frame number and not the physical address. It is also better to drop the phrase 'guest' from the name as this applies to root as well.
- Sunil
Quoted 61 lines · Sunil Muthuswamy
From: Praveen Kumar <[email protected]> Sent: Tuesday, July 27, 2021 3:41 AM
Quoted 60 lines
For Root partition the VP assist pages are pre-determined by the hypervisor. The Root kernel is not allowed to change them to different locations. And thus, we are getting below stack as in current implementation Root is trying to perform write to specific MSR.
[ 2.778197] unchecked MSR access error: WRMSR to 0x40000073 (tried to write 0x0000000145ac5001) at rIP: 0xffffffff810c1084 (native_write_msr+0x4/0x30) [ 2.784867] Call Trace: [ 2.791507] hv_cpu_init+0xf1/0x1c0 [ 2.798144] ? hyperv_report_panic+0xd0/0xd0 [ 2.804806] cpuhp_invoke_callback+0x11a/0x440 [ 2.811465] ? hv_resume+0x90/0x90 [ 2.818137] cpuhp_issue_call+0x126/0x130 [ 2.824782] __cpuhp_setup_state_cpuslocked+0x102/0x2b0 [ 2.831427] ? hyperv_report_panic+0xd0/0xd0 [ 2.838075] ? hyperv_report_panic+0xd0/0xd0 [ 2.844723] ? hv_resume+0x90/0x90 [ 2.851375] __cpuhp_setup_state+0x3d/0x90 [ 2.858030] hyperv_init+0x14e/0x410 [ 2.864689] ? enable_IR_x2apic+0x190/0x1a0 [ 2.871349] apic_intr_mode_init+0x8b/0x100 [ 2.878017] x86_late_time_init+0x20/0x30 [ 2.884675] start_kernel+0x459/0x4fb [ 2.891329] secondary_startup_64_no_verify+0xb0/0xbbSince, the hypervisor already provides the VP assist page for root partition, we need to memremap the memory from hypervisor for root kernel to use. The mapping is done in
hv_cpu_initduring bringup and is unmaped inhv_cpu_dieduring teardown.Signed-off-by Praveen Kumar <[email protected]>--- arch/x86/hyperv/hv_init.c | 61 +++++++++++++++++++++--------- arch/x86/include/asm/hyperv-tlfs.h | 9 +++++ 2 files changed, 53 insertions(+), 17 deletions(-)changelog:
v1: initial patch
v2: commit message changes, removal ofHV_MSR_APIC_ACCESS_AVAILABLE
check and addition of null check before reading the VP assist MSR for root partition
v3: added new data structure to handle VP ASSIST MSR page and done
handling inhv_cpu_initandhv_cpu_die---
arch/x86/hyperv/hv_init.c +1
-0@@ -44,6 +44,7 @@ EXPORT_SYMBOL_GPL(hv_vp_assist_page);static int hv_cpu_init (unsigned int cpu) { +union hv_vp_assist_msr_contents msr;struct hv_vp_assist_page **hvp = &hv_vp_assist_page[smp_processor_id ()];int ret;
Quoted 8 lines
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 56
Quoted 3 lines
arch/x86/hyperv/hv_init.c +0
-0@@ -54,27 +55,41 @@ static int hv_cpu_init(unsigned int cpu)if (!hv_vp_assist_page)return 0 ;Not related to this code, but I am not sure about the usefulness of this NULL check as we have already accessed this pointer above. If it was NULL, things would already blow up.
What I understood, hvp will point to "hv_vp_assist_page stack address + smp_processor_id()"
So, we are good, and this NULL check is required, as in when we de-reference the location, later in the code, it may fault. Please do correct me if my understanding is wrong here. Thanks.
Quoted 12 lines
Quoted 12 lines
arch/x86/hyperv/hv_init.c +2
-9-/* -* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's Section -* 5.2.1 "GPA Overlay Pages"). Here it must be zeroed out to make sure -* we always write the EOI MSR in hv_apic_eoi_write() *after* the -* EOI optimization is disabled in hv_cpu_die(), otherwise a CPU may -* not be stopped in the case of CPU offlining and the VM will hang. -*/ -if (!*hvp) { - *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); +if (hv_root_partition) { +/*
Quoted 9 lines
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 65
Quoted 6 lines
arch/x86/hyperv/hv_init.c +6
-0+* For Root partition we get the hypervisor provided VP ASSIST +* PAGE, instead of allocating a new page. +*/ +rdmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64); + +/* remapping to root partition address space */ Better to leave out comments that are obvious from the code.
Sure will remove the comment here. Thanks.
Quoted 22 lines
Quoted 22 lines
arch/x86/hyperv/hv_init.c +17
-2+if (!*hvp) + *hvp =memremap (msr.guest_physical_address << +HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT , +PAGE_SIZE ,MEMREMAP_WB ); + }else { +/* +* The VP ASSIST PAGE is an "overlay" page (see Hyper-V TLFS's +* Section 5.2.1 "GPA Overlay Pages"). Here it must be zeroed +* out to make sure we always write the EOI MSR in +* hv_apic_eoi_write() *after* theEOI optimization is disabled +* in hv_cpu_die(), otherwise a CPU may not be stopped in the +* case of CPU offlining and the VM will hang. +*/ +if (!*hvp) + *hvp =__vmalloc (PAGE_SIZE ,GFP_KERNEL | __GFP_ZERO); + } -if (*hvp) { -u64 val; +WARN_ON (!(*hvp));
Quoted 12 lines
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 88
Quoted 6 lines
arch/x86/hyperv/hv_init.c +3
-3- val =vmalloc_to_pfn (*hvp); - val = (val <<HV_X64_MSR_VP_ASSIST_PAGE_ADDRESS_SHIFT ) | -HV_X64_MSR_VP_ASSIST_PAGE_ENABLE ; +if (*hvp) { +if (!hv_root_partition) + msr.guest_physical_address =vmalloc_to_pfn (*hvp);It's better to move this above in the else section where we are 'vmalloc' the page. If you just check for the NULL for the page above and return if NULL, that should clean up the code as well.
Quoted 1 line
Ok, I was trying to make code cleaner, but I would have done better.
Quoted 13 lines
arch/x86/hyperv/hv_init.c static int hv_cpu_init(unsigned int cpu) · line 90
Quoted 2 lines
arch/x86/hyperv/hv_init.c +1
-1-wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , val); + msr.enable =1 ;We should also set the reserved bits to 0.
Quoted 8 lines
arch/x86/hyperv/hv_init.c +1
-1+wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE , msr.as_uint64); } -return 0 ; }@@ -170,9 +185,21 @@ static int hv_cpu_die(unsigned int cpu)
Quoted 8 lines
arch/x86/hyperv/hv_init.c static int hv_cpu_die(unsigned int cpu) · line 189
Quoted 5 lines
arch/x86/hyperv/hv_init.c +1
-1hv_common_cpu_die (cpu); -if (hv_vp_assist_page && hv_vp_assist_page[cpu]) +if (hv_vp_assist_page && hv_vp_assist_page[cpu]) {wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE ,0 );Its better to read the MSR, set the enable bit to 0 and write it back.
Ok. probably this is the same code, which is being done in hv_cpu_init just with enable bit set to 0. Let me try of a API / macro which can be used.
Quoted 14 lines
arch/x86/hyperv/hv_init.c static int hv_cpu_die(unsigned int cpu) · line 200
Quoted 11 lines
arch/x86/hyperv/hv_init.c +10
-0+if (hv_root_partition) { +/* +* For Root partition the VP ASSIST page is mapped to +* hypervisor provided page, and thus, we unmap the +* page here and nullify it, so that in future we have +* correct page address mapped in hv_cpu_init +*/ +memunmap (hv_vp_assist_page[cpu]); + hv_vp_assist_page[cpu] =NULL ; + }For the guest case, where are we freeing the page?
Yes, this I have observed,and looks like we don't want to allocate the page every-time, instead reuse the same. If someone in the list can please provide some information or history behind this, it would give some insights. But for now, I am leaving it as it is.
Quoted 13 lines
Quoted 13 lines
arch/x86/hyperv/hv_init.c +2
-0+ } +if (hv_reenlightenment_cb ==NULL )return 0 ;arch/x86/include/asm/hyperv-tlfs.h +0
-0@@ -288,6 +288,15 @@ union hv_x64_msr_hypercall_contents { } __packed; };
Quoted 11 lines
arch/x86/include/asm/hyperv-tlfs.h union hv_x64_msr_hypercall_contents { · line 296
Quoted 6 lines
arch/x86/include/asm/hyperv-tlfs.h +6
-0+union hv_vp_assist_msr_contents { +u64 as_uint64; +struct { +u64 enable:1 ; +u64 reserved:11 ; +u64 guest_physical_address:52 ;This field contains the page frame number and not the physical address. It is also better to drop the phrase 'guest' from the name as this applies to root as well.
Ack.
Quoted 1 line
- Sunil
Regards,
~Praveen.
Quoted 1 line
Quoted 14 lines
Quoted 6 lines
Quoted 6 lines
static int
hv_cpu_init(unsigned int cpu)
{arch/x86/hyperv/hv_init.c +1
-0+union hv_vp_assist_msr_contents msr;struct hv_vp_assist_page **hvp = &hv_vp_assist_page[smp_processor_id ()];int ret;Quoted 8 lines
Review static int hv_cpu_init(unsigned int cpu) · line 56
Quoted 3 lines
(quoted diff) +0
-0@@ -54,27 +55,41 @@ static int hv_cpu_init(unsigned int cpu) if (!hv_vp_assist_page) return 0;Not related to this code, but I am not sure about the usefulness of this NULL check as we have already accessed this pointer above. If it was NULL, things would already blow up.
Quoted 4 lines
What I understood, hvp will point to "
hv_vp_assist_pagestack address +smp_processor_id()"
So, we are good, and this NULL check is required, as in when we de-reference the location, later in the code, it may fault. Please do correct me if my understanding is wrong here. Thanks.
'hv_vp_assist_page' comes from the heap, there is nothing on the stack there. As I mentioned previously, if 'hv_vp_assist_page' was NULL, we would have already crashed by now, as we are accessing it above. So, the check here is useless in my opinion. But, since this is code that this patch doesn't touch, its fine to leave it as is for this patch. I was just pointing it out.
- Sunil
Wei Liu
maintainerOn Tue, Jul 27, 2021 at 04:10:44PM +0530, Praveen Kumar wrote:
[...]
arch/x86/hyperv/hv_init.c static int hv_cpu_die(unsigned int cpu) · line 190
Quoted 9 lines
arch/x86/hyperv/hv_init.c +1
-1@@ -170,9 +185,21 @@ static int hv_cpu_die(unsigned int cpu)hv_common_cpu_die (cpu); -if (hv_vp_assist_page && hv_vp_assist_page[cpu]) +if (hv_vp_assist_page && hv_vp_assist_page[cpu]) {wrmsrl (HV_X64_MSR_VP_ASSIST_PAGE ,0 );
The content of the MSR should be preserved; otherwise you hit the same fault for root kernel.
Wei.