Page MenuHomeFreeBSD

bhyveUmbrella
ActivePublic

Details

Description

bhyve reviewers group

Recent Activity

Yesterday

gusev.vitaliy_gmail.com updated the summary of D35590: bhyve: multiple devices support for suspend/checkpoint/resume.
Fri, Jun 24, 8:01 PM · bhyve
gusev.vitaliy_gmail.com requested review of D35590: bhyve: multiple devices support for suspend/checkpoint/resume.
Fri, Jun 24, 7:57 PM · bhyve

Thu, Jun 23

rew added a comment to D35454: bhyve: snapshot capsicum support.

If use one fd for all steps (save all things - ram, kernel, config to one file) - it will require completly rework saving format.

Thu, Jun 23, 11:15 PM · bhyve
jhb added inline comments to D28280: bhyve: set lpc IDs to physical values.
Thu, Jun 23, 7:03 PM · bhyve
jhb closed D26267: bhyve: snapshot impovements for 'blockif' backend.
Thu, Jun 23, 6:47 PM · bhyve
gusev.vitaliy_gmail.com added inline comments to D35575: bhyve: Support Hyper-V (base) and hyperv clock.
Thu, Jun 23, 6:40 PM · bhyve
gusev.vitaliy_gmail.com added a comment to D35454: bhyve: snapshot capsicum support.
In D35454#806691, @rew wrote:

not sold on it

My perspective about save/restore is that the limitations/file format issues described in the original commit message, 483d953a86a2507355f8287c5107dc827a0ff516, need to be resolved/discussed.

Until those issues are addressed, the save/restore feature (and any feature that builds off of it) will not make it into main.

Thu, Jun 23, 6:18 PM · bhyve
emaste added inline comments to D29262: Bhyve - Using JSON format for saving and restoring the state.
Thu, Jun 23, 5:22 PM · bhyve
mgrooms_shrew.net added a comment to D35454: bhyve: snapshot capsicum support.

Do these get bhyve closer to what you believe to be required?

Thu, Jun 23, 5:14 PM · bhyve
pmooney_pfmooney.com added inline comments to D35575: bhyve: Support Hyper-V (base) and hyperv clock.
Thu, Jun 23, 5:03 PM · bhyve
jhb accepted D26267: bhyve: snapshot impovements for 'blockif' backend.

I will work on merging this. I think it probably makes sense to do this as a couple of commits:

Thu, Jun 23, 4:53 PM · bhyve
rew added a comment to D35454: bhyve: snapshot capsicum support.

not sold on it

Thu, Jun 23, 4:49 PM · bhyve
Herald added a reviewer for D14473: userboot: add callbacks to set unrestricted guest mode: manu.

Do you still need the vm_set_unrestricted_guest callback? It seems easy to at least add that part of this patch.

Thu, Jun 23, 4:37 PM · bhyve
jhb added inline comments to D35575: bhyve: Support Hyper-V (base) and hyperv clock.
Thu, Jun 23, 4:24 PM · bhyve
afedorov added a reviewer for D35575: bhyve: Support Hyper-V (base) and hyperv clock: howard0su_gmail.com.
Thu, Jun 23, 2:55 PM · bhyve
afedorov added a reviewer for D35575: bhyve: Support Hyper-V (base) and hyperv clock: kib.
Thu, Jun 23, 2:54 PM · bhyve
gusev.vitaliy_gmail.com requested review of D35575: bhyve: Support Hyper-V (base) and hyperv clock.
Thu, Jun 23, 2:22 PM · bhyve
gusev.vitaliy_gmail.com added a comment to D35454: bhyve: snapshot capsicum support.

@rew What do you think about idea implementing "bricks/handlers" in bhyve and full logic in bhyvectl ?

Thu, Jun 23, 1:52 PM · bhyve

Tue, Jun 21

evgueni.gavrilov_itglobal.com added a comment to D35073: libvmm: add vm_close() to vmmapi library.

vm_close() is widely used at vStack platform already
it would be nice to use work hours sponsorship :shrug

Tue, Jun 21, 12:14 PM · bhyve

Sat, Jun 18

gusev.vitaliy_gmail.com added a comment to D35454: bhyve: snapshot capsicum support.
In D35454#804127, @rew wrote:

I'd rather see the snapshot functionality write it's data to a single file. A file descriptor for said file would be passed in from the nvlist via an IPC command.

In other words, bhyvectl would open the snapshot file and pass that file descriptor over the socket to bhyve.

Sat, Jun 18, 7:11 PM · bhyve

Fri, Jun 17

pmooney_pfmooney.com added a comment to D35447: vmm: save/restore pir_desc is required for posted-interrupts and VID.

I guess you meant that vmx_apicv_sync() should be called instead saving pir_desc directly ? I understand what you mean, but it will also require saving of vlapic_vtx. Current implementation in FreeBSD also requies saving vlapic_vtx, but just one field pending_prio. I am going to create review for that, once current review is finished.

Fri, Jun 17, 9:11 PM · bhyve
gusev.vitaliy_gmail.com added a reviewer for D35447: vmm: save/restore pir_desc is required for posted-interrupts and VID: pmooney_pfmooney.com.
Fri, Jun 17, 8:50 PM · bhyve
gusev.vitaliy_gmail.com added a comment to D35447: vmm: save/restore pir_desc is required for posted-interrupts and VID.

Sorry, If vlapic_set_intr_ready() is called, it will change pir_desc->pir in some condition:
I assume that it can lead that pir_desc->pir is only container that keeps a interrupt request. Therefore if suspend VM and then restore, that interrupt will be lost.

Operations like save/restore require that all vCPUs and associated emulation be held from running. In that moment pir_desc should be unchanging, and thus safe to flatten in the APIC page without losing any interrupts.

I guess you meant that vmx_apicv_sync() should be called instead saving pir_desc directly ? I understand what you mean, but it will also require saving of vlapic_vtx. Current implementation in FreeBSD also requies saving vlapic_vtx, but just one field pending_prio. I am going to create review for that, once current review is finished.

Fri, Jun 17, 8:49 PM · bhyve
pmooney_pfmooney.com added a comment to D35447: vmm: save/restore pir_desc is required for posted-interrupts and VID.

Sorry, If vlapic_set_intr_ready() is called, it will change pir_desc->pir in some condition:
I assume that it can lead that pir_desc->pir is only container that keeps a interrupt request. Therefore if suspend VM and then restore, that interrupt will be lost.

Fri, Jun 17, 8:12 PM · bhyve
gusev.vitaliy_gmail.com added a comment to D35447: vmm: save/restore pir_desc is required for posted-interrupts and VID.

But this commit should also suffer from not saved pir_desc after vmx_apicv_set_ready() is performed. Am I right?

The point of approach is to merge the contents of pir_desc into the vlapic (really the APIC page) itself. The asserted bits in pir_desc are simply requests to queue a pending interrupt in the virtual APIC. That is effectively the same as an asserted bit in the IRR fields of the APIC page. By flattening those bits into the APIC page, it eliminates the need to save PIR data independently.
s

Sorry, If vlapic_set_intr_ready() is called, it will change pir_desc->pir in some condition:

Fri, Jun 17, 7:59 PM · bhyve
pmooney_pfmooney.com added a comment to D35447: vmm: save/restore pir_desc is required for posted-interrupts and VID.

But this commit should also suffer from not saved pir_desc after vmx_apicv_set_ready() is performed. Am I right?

Fri, Jun 17, 7:50 PM · bhyve
gusev.vitaliy_gmail.com added a comment to D35447: vmm: save/restore pir_desc is required for posted-interrupts and VID.

In illumos bhyve, we chose to "flatten" queued interrupts in the PIR descriptor into the normal vlapic state, so there was only one source of truth for the IRR data:

https://github.com/illumos/illumos-gate/blob/e760f15095bdc9fa107e7c20ed2a5e4fb5865c1d/usr/src/uts/intel/io/vmm/intel/vmx.c#L3532-L3597

The sync operation is called when we need the IRR data to be up to date, like when checking for pending interrupts (from the HLT handler, or during VM entry) and when the vlapic data is about to be exported for save/restore. That way, neither source nor destination need to be explicitly aware of APICv.
Perhaps it's an approach worth considering in FreeBSD.

Fri, Jun 17, 7:09 PM · bhyve
pmooney_pfmooney.com added a comment to D35447: vmm: save/restore pir_desc is required for posted-interrupts and VID.

In illumos bhyve, we chose to "flatten" queued interrupts in the PIR descriptor into the normal vlapic state, so there was only one source of truth for the IRR data:

Fri, Jun 17, 12:01 AM · bhyve

Sat, Jun 11

rew added a comment to D35454: bhyve: snapshot capsicum support.

I'd rather see the snapshot functionality write it's data to a single file. A file descriptor for said file would be passed in from the nvlist via an IPC command.

Sat, Jun 11, 6:18 AM · bhyve

Fri, Jun 10

gusev.vitaliy_gmail.com requested review of D35454: bhyve: snapshot capsicum support.
Fri, Jun 10, 11:50 PM · bhyve
gusev.vitaliy_gmail.com requested review of D35447: vmm: save/restore pir_desc is required for posted-interrupts and VID.
Fri, Jun 10, 6:27 PM · bhyve
gusev.vitaliy_gmail.com added a comment to D26267: bhyve: snapshot impovements for 'blockif' backend.

@jhb @markj Could you look again and approve ?

Fri, Jun 10, 11:24 AM · bhyve
gusev.vitaliy_gmail.com updated the summary of D35073: libvmm: add vm_close() to vmmapi library.
Fri, Jun 10, 11:22 AM · bhyve
gusev.vitaliy_gmail.com updated the summary of D26267: bhyve: snapshot impovements for 'blockif' backend.
Fri, Jun 10, 10:58 AM · bhyve
gusev.vitaliy_gmail.com abandoned D26266: bhyve: snapshot fix for ahci-cd.
Fri, Jun 10, 10:23 AM · bhyve
gusev.vitaliy_gmail.com added inline comments to D26267: bhyve: snapshot impovements for 'blockif' backend.
Fri, Jun 10, 10:21 AM · bhyve
gusev.vitaliy_gmail.com added a comment to D26266: bhyve: snapshot fix for ahci-cd.

Close that review in favour of D26267

Fri, Jun 10, 10:17 AM · bhyve
gusev.vitaliy_gmail.com updated the diff for D26267: bhyve: snapshot impovements for 'blockif' backend.

Corrected according to comments and suggestions.

Fri, Jun 10, 10:15 AM · bhyve

Thu, Jun 9

markj accepted D35442: vmm: Cherry pick illumos commit '13361 bhyve should mask RDT cpuid info'.
Thu, Jun 9, 2:14 PM · bhyve
gusev.vitaliy_gmail.com requested review of D35442: vmm: Cherry pick illumos commit '13361 bhyve should mask RDT cpuid info'.
Thu, Jun 9, 2:11 PM · bhyve
markj closed D35350: vmm: move bumping VMEXIT_USERSPACE stat to the right place.
Thu, Jun 9, 1:09 PM · bhyve

Wed, Jun 8

gusev.vitaliy_gmail.com added a comment to D35350: vmm: move bumping VMEXIT_USERSPACE stat to the right place.

@markj Could you commit it?

Wed, Jun 8, 8:27 PM · bhyve
gusev.vitaliy_gmail.com added a comment to D35073: libvmm: add vm_close() to vmmapi library.

@jhb up? Could you commit it?

Wed, Jun 8, 8:26 PM · bhyve

Mon, May 30

markj accepted D35350: vmm: move bumping VMEXIT_USERSPACE stat to the right place.
Mon, May 30, 2:32 PM · bhyve
gusev.vitaliy_gmail.com requested review of D35350: vmm: move bumping VMEXIT_USERSPACE stat to the right place.
Mon, May 30, 1:47 PM · bhyve
manu closed D35253: vmm: add tunable to trap WBINVD.
Mon, May 30, 9:19 AM · bhyve
manu closed D34465: bhyve: use bhyve_config for SMBIOS strings.
Mon, May 30, 9:19 AM · bhyve

May 23 2022

jhb accepted D35253: vmm: add tunable to trap WBINVD.
May 23 2022, 6:12 PM · bhyve

May 21 2022

c.koehne_beckhoff.com added inline comments to D35253: vmm: add tunable to trap WBINVD.
May 21 2022, 2:38 PM · bhyve

May 20 2022

jhb added inline comments to D35253: vmm: add tunable to trap WBINVD.
May 20 2022, 5:35 PM · bhyve