Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Low] mshv: missing Fixes trailer for commit 621191d709b1
It also noticed 1 related pre-existing issue this patch does not
introduce:
- [High] mshv: reclaim lock inversion in hv_do_map_gpa_hcall() under
  mreg_mutex [1]
--

commit f6f04457a79d386b8628402d94d80fb4ab5e158c
Author: Mukesh R <[email protected]>

> mshv: Handle completions on repeat hypercall failures
>
> When a repeat hypercall fails, it could still have completed some of the
> entries requested. Fix places where completions are not checked in case
> of repeat hypercall failures. Also, add error prints in couple places
> with correct partial completion counts when repeat hypercall fails.
>
> Signed-off-by: Mukesh R <[email protected]>

[Severity: Low]
All four modified helpers were introduced with the status check placed
before accounting for hv_repcomp(status).  Should this include:

  Fixes: 621191d709b1 ("Drivers: hv: Introduce mshv_root module to expose 
/dev/mshv to VMMs")

> diff --git a/drivers/hv/mshv_root_hv_call.c b/drivers/hv/mshv_root_hv_call.c
> index e761addda50e1..6261686c36382 100644
> --- a/drivers/hv/mshv_root_hv_call.c
> +++ b/drivers/hv/mshv_root_hv_call.c
> @@ -239,6 +239,7 @@ static int hv_do_map_gpa_hcall(u64 partition_id, u64 gfn, 
> u64 page_struct_count,
>               local_irq_restore(irq_flags);
>  
>               completed = hv_repcomp(status);
> +             done += completed;
>  
>               if (hv_result_needs_memory(status)) {
>                       ret = hv_deposit_memory(partition_id, status);

[Severity: High]
This problem wasn't introduced by this patch, but can calling
hv_deposit_memory() here cause a reclaim lock inversion on
region->mreg_mutex [1]?

mshv_region_range_fault() and mshv_region_interval_invalidate() call
mshv_region_remap_pages() -> ... -> hv_do_map_gpa_hcall() while holding
region->mreg_mutex.  When hv_result_needs_memory(status) is true,
hv_deposit_memory() allocates pages with GFP_KERNEL via
hv_call_deposit_pages() and hv_alloc_dep_pages(), which can enter direct
reclaim and invoke the MMU interval notifier callback on the same region:

mshv_region_range_fault()
  mutex_lock(&region->mreg_mutex)
  mshv_region_remap_pages()
    ...
      hv_do_map_gpa_hcall()
        hv_deposit_memory()
          ...
            hv_call_deposit_pages()
              get_zeroed_page(GFP_KERNEL)
                <direct reclaim>
                  mmu_notifier_invalidate_range_start()
                    mshv_region_interval_invalidate()
                      mutex_lock(&region->mreg_mutex)  <-- deadlock

[1] https://sashiko.dev/#/bug/linux-5f8ad8fe-ef3b-459b-a784-e85fb87d7746

-- 
Sashiko AI review ยท 
https://sashiko.dev/#/patchset/[email protected]?part=1

Reply via email to