Re: [Qemu-devel] [PATCH v2] translate-all: Bugfix for user-mode self-modifying code in 2 page long TB

[Sergey Fedorov <sergey.fedorov@linaro.org>]

On 06/07/16 15:01, Стас Шмаров wrote:
> May be this can be done faster.
> After invalidation of current TB, we know next executed TB will be 1
> instruction long and will write to same address, and will cause
> segfault. In sigfault handler write protection to this page will be
> disabled.
> So, we can retranslate TB, and remove protection before execution of
> new TB. It will be one signal less.

You mean not to return 2 immediately but rather to remember the desired
return value and keep invalidating guest pages, then go ahead and call
mprotect() to enable writes to the host page?

Thanks,
Sergey

>
> 2016-07-06 13:21 GMT+03:00 Sergey Fedorov <sergey.fedorov@linaro.org
> <mailto:sergey.fedorov@linaro.org>>:
>
>     On 06/07/16 10:54, Stanislav Shmarov wrote:
>     > In user-mode emulation Translation Block can consist of 2 guest
>     pages.
>     > In that case QEMU also mprotects 2 host pages that are dedicated for
>     > guest memory, containing instructions. QEMU detects
>     self-modifying code
>     > with SEGFAULT signal processing.
>     >
>     > In case if instruction in 1st page is modifying memory of 2nd
>     > page (or vice versa) QEMU will mark 2nd page with PAGE_WRITE,
>     > invalidate TB, generate new TB contatining 1 guest instruction and
>     > exit to CPU loop. QEMU won't call mprotect, and new TB will cause
>     > same SEGFAULT. Page will have both PAGE_WRITE_ORG and PAGE_WRITE
>     > flags, so QEMU will handle the signal as guest binary problem,
>     > and exit with guest SEGFAULT.
>     >
>     > Solution is retranslate TB before marking pages as PAGE_WRITE,
>     > and remove protection with mprotect on second SEGFAULT.
>     >
>     > Signed-off-by: Stanislav Shmarov <snarpix@gmail.com
>     <mailto:snarpix@gmail.com>>
>
>     Reviewed-by: Sergey Fedorov <sergey.fedorov@linaro.org
>     <mailto:sergey.fedorov@linaro.org>>
>
>     > ---
>     >   v2: Moved setting PAGE_WRITE flag to separte loop, to cover cases,
>     >       pointed by Sergey Fedorov.
>     >
>     >  translate-all.c | 17 +++++++++++------
>     >  1 file changed, 11 insertions(+), 6 deletions(-)
>     >
>     > diff --git a/translate-all.c b/translate-all.c
>     > index eaa95e4..fb3743f 100644
>     > --- a/translate-all.c
>     > +++ b/translate-all.c
>     > @@ -2020,13 +2020,8 @@ int page_unprotect(target_ulong address,
>     uintptr_t pc)
>     >          host_start = address & qemu_host_page_mask;
>     >          host_end = host_start + qemu_host_page_size;
>     >
>     > -        prot = 0;
>     >          for (addr = host_start ; addr < host_end ; addr +=
>     TARGET_PAGE_SIZE) {
>     > -            p = page_find(addr >> TARGET_PAGE_BITS);
>     > -            p->flags |= PAGE_WRITE;
>     > -            prot |= p->flags;
>     > -
>     > -            /* and since the content will be modified, we must
>     invalidate
>     > +            /* Since the content will be modified, we must
>     invalidate
>     >                 the corresponding translated code. */
>     >              if (tb_invalidate_phys_page(addr, pc)) {
>     >                  mmap_unlock();
>     > @@ -2036,6 +2031,16 @@ int page_unprotect(target_ulong address,
>     uintptr_t pc)
>     >              tb_invalidate_check(addr);
>     >  #endif
>     >          }
>     > +
>     > +        /* If we got here, current TB have been retranslated
>     (in case of
>     > +         * self-modifying code), now it's safe to remove page
>     protection.
>     > +         */
>     > +        prot = 0;
>     > +        for (addr = host_start ; addr < host_end ; addr +=
>     TARGET_PAGE_SIZE) {
>     > +            p = page_find(addr >> TARGET_PAGE_BITS);
>     > +            p->flags |= PAGE_WRITE;
>     > +            prot |= p->flags;
>     > +        }
>     >          mprotect((void *)g2h(host_start), qemu_host_page_size,
>     >                   prot & PAGE_BITS);
>     >
>
>