Re: Re: [PATCH 0/3] recover hardware corrupted page by virtio balloon
On 6/1/22 15:59, David Hildenbrand wrote: On 01.06.22 04:17, zhenwei pi wrote: On 5/31/22 12:08, Jue Wang wrote: On Mon, May 30, 2022 at 8:49 AM Peter Xu wrote: On Mon, May 30, 2022 at 07:33:35PM +0800, zhenwei pi wrote: A VM uses RAM of 2M huge page. Once a MCE(@HVAy in [HVAx,HVAz)) occurs, the 2M([HVAx,HVAz)) of hypervisor becomes unaccessible, but the guest poisons 4K (@GPAy in [GPAx, GPAz)) only, it may hit another 511 MCE ([GPAx, GPAz) except GPAy). This is the worse case, so I want to add '__le32 corrupted_pages' in struct virtio_balloon_config, it is used in the next step: reporting 512 * 4K 'corrupted_pages' to the guest, the guest has a chance to isolate the other 511 pages ahead of time. And the guest actually loses 2M, fixing 512*4K seems to help significantly. It sounds hackish to teach a virtio device to assume one page will always be poisoned in huge page granule. That's only a limitation to host kernel not virtio itself. E.g. there're upstream effort ongoing with enabling doublemap on hugetlbfs pages so hugetlb pages can be mapped in 4k with it. It provides potential possibility to do page poisoning with huge pages in 4k too. When that'll be ready the assumption can go away, and that does sound like a better approach towards this problem. +1. A hypervisor should always strive to minimize the guest memory loss. The HugeTLB double mapping enlightened memory poisoning behavior (only poison 4K out of a 2MB huge page and 4K in guest) is a much better solution here. To be completely transparent, it's not _strictly_ required to poison the page (whatever the granularity it is) on the host side, as long as the following are true: 1. A hypervisor can emulate the _minimized_ (e.g., 4K) the poison to the guest. 2. The host page with the UC error is "isolated" (could be PG_HWPOISON or in some other way) and prevented from being reused by other processes. For #2, PG_HWPOISON and HugeTLB double mapping enlightened memory poisoning is a good solution. I assume when talking about "the performance memory drops a lot", you imply that this patch set can mitigate that performance drop? But why do you see a performance drop? Because we might lose some possible THP candidates (in the host or the guest) and you want to plug does holes? I assume you'll see a performance drop simply because poisoning memory is expensive, including migrating pages around on CE. If you have some numbers to share, especially before/after this change, that would be great. The CE storm leads 2 problems I have even seen: 1, the memory bandwidth slows down to 10%~20%, and the cycles per instruction of CPU increases a lot. 2, the THR (/proc/interrupts) interrupts frequently, the CPU has to use a lot time to handle IRQ. Totally no good knowledge on CMCI, but if 2) is true then I'm wondering whether it's necessary to handle the interrupts that frequently. When I was reading the Intel CMCI vector handler I stumbled over this comment: /* * The interrupt handler. This is called on every event. * Just call the poller directly to log any events. * This could in theory increase the threshold under high load, * but doesn't for now. */ static void intel_threshold_interrupt(void) I think that matches with what I was thinking.. I mean for 2) not sure whether it can be seen as a CMCI problem and potentially can be optimized by adjust the cmci threshold dynamically. The CE storm caused performance drop is caused by the extra cycles spent by the ECC steps in memory controller, not in CMCI handling. This is observed in the Google fleet as well. A good solution is to monitor the CE rate closely in user space via /dev/mcelog and migrate all VMs to another host once the CE rate exceeds some threshold. CMCI is a _background_ interrupt that is not handled in the process execution context and its handler is setup to switch to poll (1 / 5 min) mode if there are more than ~ a dozen CEs reported via CMCI per second. -- Peter Xu Hi, Andrew, David, Naoya According to the suggestions, I'd give up the improvement of memory failure on huge page in this series. Is it worth recovering corrupted pages for the guest kernel? I'd follow your decision. Well, as I said, I am not sure if we really need/want this for a handful of 4k poisoned pages in a VM. As I suspected, doing so might primarily be interesting for some sort of de-fragmentation (allow again a higher order page to be placed at the affected PFNs), not because of the slight reduction of available memory. A simple VM reboot would get the job similarly done. Sure, Let's drop this idea. Thanks to all for the suggestions. Hi, Naoya It seems that memory failure notifier is not required currently, so I'll not push the next version of: [PATCH 1/3] memory-failure: Introduce memory failure notifier [PATCH 2/3] mm/memory-failure.c: support reset PTE during unpoison Thanks you for review work! As the poisoning refcount code is already a bit shaky as I
Re: Re: Re: [PATCH 0/3] recover hardware corrupted page by virtio balloon
On 5/31/22 12:08, Jue Wang wrote: On Mon, May 30, 2022 at 8:49 AM Peter Xu wrote: On Mon, May 30, 2022 at 07:33:35PM +0800, zhenwei pi wrote: A VM uses RAM of 2M huge page. Once a MCE(@HVAy in [HVAx,HVAz)) occurs, the 2M([HVAx,HVAz)) of hypervisor becomes unaccessible, but the guest poisons 4K (@GPAy in [GPAx, GPAz)) only, it may hit another 511 MCE ([GPAx, GPAz) except GPAy). This is the worse case, so I want to add '__le32 corrupted_pages' in struct virtio_balloon_config, it is used in the next step: reporting 512 * 4K 'corrupted_pages' to the guest, the guest has a chance to isolate the other 511 pages ahead of time. And the guest actually loses 2M, fixing 512*4K seems to help significantly. It sounds hackish to teach a virtio device to assume one page will always be poisoned in huge page granule. That's only a limitation to host kernel not virtio itself. E.g. there're upstream effort ongoing with enabling doublemap on hugetlbfs pages so hugetlb pages can be mapped in 4k with it. It provides potential possibility to do page poisoning with huge pages in 4k too. When that'll be ready the assumption can go away, and that does sound like a better approach towards this problem. +1. A hypervisor should always strive to minimize the guest memory loss. The HugeTLB double mapping enlightened memory poisoning behavior (only poison 4K out of a 2MB huge page and 4K in guest) is a much better solution here. To be completely transparent, it's not _strictly_ required to poison the page (whatever the granularity it is) on the host side, as long as the following are true: 1. A hypervisor can emulate the _minimized_ (e.g., 4K) the poison to the guest. 2. The host page with the UC error is "isolated" (could be PG_HWPOISON or in some other way) and prevented from being reused by other processes. For #2, PG_HWPOISON and HugeTLB double mapping enlightened memory poisoning is a good solution. I assume when talking about "the performance memory drops a lot", you imply that this patch set can mitigate that performance drop? But why do you see a performance drop? Because we might lose some possible THP candidates (in the host or the guest) and you want to plug does holes? I assume you'll see a performance drop simply because poisoning memory is expensive, including migrating pages around on CE. If you have some numbers to share, especially before/after this change, that would be great. The CE storm leads 2 problems I have even seen: 1, the memory bandwidth slows down to 10%~20%, and the cycles per instruction of CPU increases a lot. 2, the THR (/proc/interrupts) interrupts frequently, the CPU has to use a lot time to handle IRQ. Totally no good knowledge on CMCI, but if 2) is true then I'm wondering whether it's necessary to handle the interrupts that frequently. When I was reading the Intel CMCI vector handler I stumbled over this comment: /* * The interrupt handler. This is called on every event. * Just call the poller directly to log any events. * This could in theory increase the threshold under high load, * but doesn't for now. */ static void intel_threshold_interrupt(void) I think that matches with what I was thinking.. I mean for 2) not sure whether it can be seen as a CMCI problem and potentially can be optimized by adjust the cmci threshold dynamically. The CE storm caused performance drop is caused by the extra cycles spent by the ECC steps in memory controller, not in CMCI handling. This is observed in the Google fleet as well. A good solution is to monitor the CE rate closely in user space via /dev/mcelog and migrate all VMs to another host once the CE rate exceeds some threshold. CMCI is a _background_ interrupt that is not handled in the process execution context and its handler is setup to switch to poll (1 / 5 min) mode if there are more than ~ a dozen CEs reported via CMCI per second. -- Peter Xu Hi, Andrew, David, Naoya According to the suggestions, I'd give up the improvement of memory failure on huge page in this series. Is it worth recovering corrupted pages for the guest kernel? I'd follow your decision. -- zhenwei pi ___ Virtualization mailing list Virtualization@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/virtualization
Re: Re: [PATCH 0/3] recover hardware corrupted page by virtio balloon
On Mon, May 30, 2022 at 07:33:35PM +0800, zhenwei pi wrote: > A VM uses RAM of 2M huge page. Once a MCE(@HVAy in [HVAx,HVAz)) occurs, the > 2M([HVAx,HVAz)) of hypervisor becomes unaccessible, but the guest poisons 4K > (@GPAy in [GPAx, GPAz)) only, it may hit another 511 MCE ([GPAx, GPAz) > except GPAy). This is the worse case, so I want to add > '__le32 corrupted_pages' in struct virtio_balloon_config, it is used in the > next step: reporting 512 * 4K 'corrupted_pages' to the guest, the guest has > a chance to isolate the other 511 pages ahead of time. And the guest > actually loses 2M, fixing 512*4K seems to help significantly. It sounds hackish to teach a virtio device to assume one page will always be poisoned in huge page granule. That's only a limitation to host kernel not virtio itself. E.g. there're upstream effort ongoing with enabling doublemap on hugetlbfs pages so hugetlb pages can be mapped in 4k with it. It provides potential possibility to do page poisoning with huge pages in 4k too. When that'll be ready the assumption can go away, and that does sound like a better approach towards this problem. > > > > > I assume when talking about "the performance memory drops a lot", you > > imply that this patch set can mitigate that performance drop? > > > > But why do you see a performance drop? Because we might lose some > > possible THP candidates (in the host or the guest) and you want to plug > > does holes? I assume you'll see a performance drop simply because > > poisoning memory is expensive, including migrating pages around on CE. > > > > If you have some numbers to share, especially before/after this change, > > that would be great. > > > > The CE storm leads 2 problems I have even seen: > 1, the memory bandwidth slows down to 10%~20%, and the cycles per > instruction of CPU increases a lot. > 2, the THR (/proc/interrupts) interrupts frequently, the CPU has to use a > lot time to handle IRQ. Totally no good knowledge on CMCI, but if 2) is true then I'm wondering whether it's necessary to handle the interrupts that frequently. When I was reading the Intel CMCI vector handler I stumbled over this comment: /* * The interrupt handler. This is called on every event. * Just call the poller directly to log any events. * This could in theory increase the threshold under high load, * but doesn't for now. */ static void intel_threshold_interrupt(void) I think that matches with what I was thinking.. I mean for 2) not sure whether it can be seen as a CMCI problem and potentially can be optimized by adjust the cmci threshold dynamically. -- Peter Xu ___ Virtualization mailing list Virtualization@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/virtualization
Re: Re: [PATCH 0/3] recover hardware corrupted page by virtio balloon
On 5/30/22 15:41, David Hildenbrand wrote: On 27.05.22 08:32, zhenwei pi wrote: On 5/27/22 02:37, Peter Xu wrote: On Wed, May 25, 2022 at 01:16:34PM -0700, Jue Wang wrote: The hypervisor _must_ emulate poisons identified in guest physical address space (could be transported from the source VM), this is to prevent silent data corruption in the guest. With a paravirtual approach like this patch series, the hypervisor can clear some of the poisoned HVAs knowing for certain that the guest OS has isolated the poisoned page. I wonder how much value it provides to the guest if the guest and workload are _not_ in a pressing need for the extra KB/MB worth of memory. I'm curious the same on how unpoisoning could help here. The reasoning behind would be great material to be mentioned in the next cover letter. Shouldn't we consider migrating serious workloads off the host already where there's a sign of more severe hardware issues, instead? Thanks, I'm maintaining 1000,000+ virtual machines, from my experience: UE is quite unusual and occurs randomly, and I did not hit UE storm case in the past years. The memory also has no obvious performance drop after hitting UE. I hit several CE storm case, the performance memory drops a lot. But I can't find obvious relationship between UE and CE. So from the point of my view, to fix the corrupted page for VM seems good enough. And yes, unpoisoning several pages does not help significantly, but it is still a chance to make the virtualization better. I'm curious why we should care about resurrecting a handful of poisoned pages in a VM. The cover letter doesn't touch on that. IOW, I'm missing the motivation why we should add additional code+complexity to unpoison pages at all. If we're talking about individual 4k pages, it's certainly sub-optimal, but does it matter in practice? I could understand if we're losing megabytes of memory. But then, I assume the workload might be seriously harmed either way already? Yes, resurrecting a handful of poisoned pages does not help significantly. And, in some ways, it seems nice to have. :D A VM uses RAM of 2M huge page. Once a MCE(@HVAy in [HVAx,HVAz)) occurs, the 2M([HVAx,HVAz)) of hypervisor becomes unaccessible, but the guest poisons 4K (@GPAy in [GPAx, GPAz)) only, it may hit another 511 MCE ([GPAx, GPAz) except GPAy). This is the worse case, so I want to add '__le32 corrupted_pages' in struct virtio_balloon_config, it is used in the next step: reporting 512 * 4K 'corrupted_pages' to the guest, the guest has a chance to isolate the other 511 pages ahead of time. And the guest actually loses 2M, fixing 512*4K seems to help significantly. I assume when talking about "the performance memory drops a lot", you imply that this patch set can mitigate that performance drop? But why do you see a performance drop? Because we might lose some possible THP candidates (in the host or the guest) and you want to plug does holes? I assume you'll see a performance drop simply because poisoning memory is expensive, including migrating pages around on CE. If you have some numbers to share, especially before/after this change, that would be great. The CE storm leads 2 problems I have even seen: 1, the memory bandwidth slows down to 10%~20%, and the cycles per instruction of CPU increases a lot. 2, the THR (/proc/interrupts) interrupts frequently, the CPU has to use a lot time to handle IRQ. But no corrupted page occurs. Migrating VM to another healthy host seems a good choice. This patch does not handle CE storm case. -- zhenwei pi ___ Virtualization mailing list Virtualization@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/virtualization
Re: Re: [PATCH 0/3] recover hardware corrupted page by virtio balloon
On 5/27/22 02:37, Peter Xu wrote: On Wed, May 25, 2022 at 01:16:34PM -0700, Jue Wang wrote: The hypervisor _must_ emulate poisons identified in guest physical address space (could be transported from the source VM), this is to prevent silent data corruption in the guest. With a paravirtual approach like this patch series, the hypervisor can clear some of the poisoned HVAs knowing for certain that the guest OS has isolated the poisoned page. I wonder how much value it provides to the guest if the guest and workload are _not_ in a pressing need for the extra KB/MB worth of memory. I'm curious the same on how unpoisoning could help here. The reasoning behind would be great material to be mentioned in the next cover letter. Shouldn't we consider migrating serious workloads off the host already where there's a sign of more severe hardware issues, instead? Thanks, I'm maintaining 1000,000+ virtual machines, from my experience: UE is quite unusual and occurs randomly, and I did not hit UE storm case in the past years. The memory also has no obvious performance drop after hitting UE. I hit several CE storm case, the performance memory drops a lot. But I can't find obvious relationship between UE and CE. So from the point of my view, to fix the corrupted page for VM seems good enough. And yes, unpoisoning several pages does not help significantly, but it is still a chance to make the virtualization better. -- zhenwei pi ___ Virtualization mailing list Virtualization@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/virtualization