qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Steven Sistare <steven.sistare@oracle.com>
To: "Cédric Le Goater" <clg@redhat.com>,
	"Zhenzhong Duan" <zhenzhong.duan@intel.com>,
	qemu-devel@nongnu.org
Cc: alex.williamson@redhat.com, eric.auger@redhat.com, chao.p.peng@intel.com
Subject: Re: [PATCH 3/4] vfio/container: Fix potential SIGSEGV when recover from unmap-all-vaddr failure
Date: Thu, 26 Jun 2025 08:53:23 -0400	[thread overview]
Message-ID: <82566d3e-edd9-4860-8fcf-c676fd112bbe@oracle.com> (raw)
In-Reply-To: <9479b426-c725-4b8c-baa9-246d0d7d1fa1@redhat.com>

On 6/24/2025 12:54 PM, Cédric Le Goater wrote:
> On 6/23/25 12:22, Zhenzhong Duan wrote:
>> cpr.saved_dma_map isn't initialized in source qemu which lead to vioc->dma_map
>> assigned a NULL value, this will trigger SIGSEGV.
> 
> I don't understand the scenario. Could you please explain more ?

Thank you Zhenzhong, I see the bug.

CPR overrides then restores dma_map in both outgoing and incoming QEMU, for
different reasons. But it only sets saved_dma_map in the target.

So, a more future-proof fix IMO is to always set saved_dma_map:

@@ -182,9 +182,9 @@ bool vfio_legacy_cpr_register_container(VFIOContainer *conta
      vmstate_register(NULL, -1, &vfio_container_vmstate, container);

      /* During incoming CPR, divert calls to dma_map. */
+    VFIOIOMMUClass *vioc = VFIO_IOMMU_GET_CLASS(bcontainer);
+    container->cpr.saved_dma_map = vioc->dma_map;
      if (cpr_is_incoming()) {
-        VFIOIOMMUClass *vioc = VFIO_IOMMU_GET_CLASS(bcontainer);
-        container->cpr.saved_dma_map = vioc->dma_map;
          vioc->dma_map = vfio_legacy_cpr_dma_map;
      }

- Steve

>> Fix it by save and restore vioc->dma_map locally.
> 
> Steve, this is cpr territory. Is it still an issue with the rest
> of the patches applied ?
>  
> Thanks,
> 
> C.
> 
>> Fixes: eba1f657cbb1 ("vfio/container: recover from unmap-all-vaddr failure")
>> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
>> ---
>>   include/hw/vfio/vfio-cpr.h | 8 +++++---
>>   hw/vfio/cpr-legacy.c       | 3 ++-
>>   2 files changed, 7 insertions(+), 4 deletions(-)
>>
>> diff --git a/include/hw/vfio/vfio-cpr.h b/include/hw/vfio/vfio-cpr.h
>> index 8bf85b9f4e..aef542e93c 100644
>> --- a/include/hw/vfio/vfio-cpr.h
>> +++ b/include/hw/vfio/vfio-cpr.h
>> @@ -16,14 +16,16 @@ struct VFIOContainer;
>>   struct VFIOContainerBase;
>>   struct VFIOGroup;
>> +typedef int (*DMA_MAP_FUNC)(const struct VFIOContainerBase *bcontainer,
>> +                            hwaddr iova, ram_addr_t size, void *vaddr,
>> +                            bool readonly, MemoryRegion *mr);
>> +
>>   typedef struct VFIOContainerCPR {
>>       Error *blocker;
>>       bool vaddr_unmapped;
>>       NotifierWithReturn transfer_notifier;
>>       MemoryListener remap_listener;
>> -    int (*saved_dma_map)(const struct VFIOContainerBase *bcontainer,
>> -                         hwaddr iova, ram_addr_t size,
>> -                         void *vaddr, bool readonly, MemoryRegion *mr);
>> +    DMA_MAP_FUNC saved_dma_map;
>>   } VFIOContainerCPR;
>>   typedef struct VFIODeviceCPR {
>> diff --git a/hw/vfio/cpr-legacy.c b/hw/vfio/cpr-legacy.c
>> index a84c3247b7..100a8db74d 100644
>> --- a/hw/vfio/cpr-legacy.c
>> +++ b/hw/vfio/cpr-legacy.c
>> @@ -148,6 +148,7 @@ static int vfio_cpr_fail_notifier(NotifierWithReturn *notifier,
>>            */
>>           VFIOIOMMUClass *vioc = VFIO_IOMMU_GET_CLASS(bcontainer);
>> +        DMA_MAP_FUNC saved_dma_map = vioc->dma_map;
>>           vioc->dma_map = vfio_legacy_cpr_dma_map;
>>           container->cpr.remap_listener = (MemoryListener) {
>> @@ -158,7 +159,7 @@ static int vfio_cpr_fail_notifier(NotifierWithReturn *notifier,
>>                                    bcontainer->space->as);
>>           memory_listener_unregister(&container->cpr.remap_listener);
>>           container->cpr.vaddr_unmapped = false;
>> -        vioc->dma_map = container->cpr.saved_dma_map;
>> +        vioc->dma_map = saved_dma_map;
>>       }
>>       return 0;
>>   }
> 



  reply	other threads:[~2025-06-26 12:54 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-06-23 10:22 [PATCH 0/4] Some trivial live update fixes Zhenzhong Duan
2025-06-23 10:22 ` [PATCH 1/4] vfio/container: Fix SIGSEGV when open container file fails Zhenzhong Duan
2025-06-24  8:53   ` Cédric Le Goater
2025-06-24  9:55   ` Cédric Le Goater
2025-06-23 10:22 ` [PATCH 2/4] vfio/container: fails mdev hotplug if add migration blocker failed Zhenzhong Duan
2025-06-24  9:21   ` Cédric Le Goater
2025-06-25 21:08     ` Cédric Le Goater
2025-06-26  1:43       ` Duan, Zhenzhong
2025-06-23 10:22 ` [PATCH 3/4] vfio/container: Fix potential SIGSEGV when recover from unmap-all-vaddr failure Zhenzhong Duan
2025-06-24 16:54   ` Cédric Le Goater
2025-06-26 12:53     ` Steven Sistare [this message]
2025-06-26 13:04       ` Steven Sistare
2025-06-23 10:22 ` [PATCH 4/4] vfio/container: Fix vfio_container_post_load() Zhenzhong Duan
2025-06-26 13:21   ` Steven Sistare

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=82566d3e-edd9-4860-8fcf-c676fd112bbe@oracle.com \
    --to=steven.sistare@oracle.com \
    --cc=alex.williamson@redhat.com \
    --cc=chao.p.peng@intel.com \
    --cc=clg@redhat.com \
    --cc=eric.auger@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=zhenzhong.duan@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).