Re: [Qemu-devel] [PATCH v2 10/23] RISC-V: Hold rcu_read_lock when accessing memory

[Philippe Mathieu-Daudé]

Hi Michael,

On 03/09/2018 05:12 AM, Michael Clark wrote:
> From reading other code that accesses memory regions directly,
> it appears that the rcu_read_lock needs to be held. Note: the
> original code for accessing RAM directly was added because
> there is no other way to use atomic_cmpxchg on guest physical
> address space.
> 
> Cc: Sagar Karandikar <address@hidden>
> Cc: Bastian Koppelmann <address@hidden>
> CC: Stefan O'Rear <address@hidden>
> Signed-off-by: Michael Clark <address@hidden>
> Signed-off-by: Palmer Dabbelt <address@hidden>
> ---
>  target/riscv/helper.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/target/riscv/helper.c b/target/riscv/helper.c
> index 02cbcea..228933c 100644
> --- a/target/riscv/helper.c
> +++ b/target/riscv/helper.c
> @@ -209,6 +209,7 @@ restart:
>                     as the PTE is no longer valid */
>                  MemoryRegion *mr;
>                  hwaddr l = sizeof(target_ulong), addr1;
> +                rcu_read_lock();
>                  mr = address_space_translate(cs->as, pte_addr,
>                      &addr1, &l, false);
>                  if (memory_access_is_direct(mr, true)) {
> @@ -222,16 +223,19 @@ restart:
>                      target_ulong old_pte =
>                          atomic_cmpxchg(pte_pa, pte, updated_pte);
>                      if (old_pte != pte) {
> +                        rcu_read_unlock();
>                          goto restart;
>                      } else {
>                          pte = updated_pte;
>                      }
>  #endif
>                  } else {
> +                    rcu_read_unlock();
>                      /* misconfigured PTE in ROM (AD bits are not preset) or
>                       * PTE is in IO space and can't be updated atomically */
>                      return TRANSLATE_FAIL;
>                  }
> +                rcu_read_unlock();

Can you refactor to have a unique pair of lock/unlock?
This would be less bug-prone.

Thanks,

Phil.

>              }
>  
>              /* for superpage mappings, make a fake leaf PTE for the TLB's
>