Re: [PATCH] tcg/tci: Fix TCI on hppa host and update TCI test matrix

[Philippe Mathieu-Daudé]

Cc'ing Akihiko for commit a1eaa6281f.

On 7/1/24 08:19, Helge Deller wrote:
> Update the TCI interpreter test matrix for big-endian hosts with
> big- (hppa, hppa64) and little-endian (x86,x96-64) targets.
> I used native ppc64 and hppa hosts for those tests.
>
> Starting TCI on a hppa host crashed immediately, because hppa is
> the only archive left where the stack grows upwards.
> Write-protecting the stack guard page at the top of the stack
> fixes the crash.

Fixes: a1eaa6281f ("util: Delete checks for old host definitions")

> Signed-off-by: Helge Deller <deller@gmx.de>
>
> diff --git a/tcg/tci/README b/tcg/tci/README
> index 4a8b5b5401..0c1e50779e 100644
> --- a/tcg/tci/README
> +++ b/tcg/tci/README
> @@ -72,16 +72,16 @@ host and target with same or different endianness.
>               | host (le)                     host (be)
>               | 32             64             32             64
>   ------------+------------------------------------------------------------
> -target (le) | s0, u0         s1, u1         s?, u?         s?, u?
> +target (le) | s0, u0         s1, u1         s2, u?         s2, u?
>   32 bit      |
>               |
> -target (le) | sc, uc         s1, u1         s?, u?         s?, u?
> +target (le) | sc, uc         s1, u1         s2, u?         s2, u?
>   64 bit      |
>               |
> -target (be) | sc, u0         sc, uc         s?, u?         s?, u?
> +target (be) | sc, u0         sc, uc         s2, u?         s2, u?
>   32 bit      |
>               |
> -target (be) | sc, uc         sc, uc         s?, u?         s?, u?
> +target (be) | sc, uc         sc, uc         s?, u?         s2, u?
>   64 bit      |
>               |
>   
> @@ -110,6 +115,10 @@ u1 = linux-user-test works
>     A cross compiled QEMU for ppc host works at least partially:
>     i386-linux-user/qemu-i386 can run a simple hello-world program
>     (tested in a ppc emulation).
> +  The big-endian tests were run on native hppa (big-endian, 32-bit) and
> +  ppc64 (big-endian, 64-bit) machines. Tested target machines were
> +  x86 and x86-64 (little-endian, debian install ISO) and 32- and 64-bit
> +  big-endian hppa (NetBSD and Debian install ISOs).
>   
>   * Some TCG opcodes are either missing in the code generator and/or
>     in the interpreter. These opcodes raise a runtime exception, so it is
> diff --git a/util/oslib-posix.c b/util/oslib-posix.c
> index e86fd64e09..e378b71641 100644
> --- a/util/oslib-posix.c
> +++ b/util/oslib-posix.c
> @@ -585,11 +585,8 @@ char *qemu_get_pid_name(pid_t pid)
>   
>   void *qemu_alloc_stack(size_t *sz)
>   {
> -    void *ptr;
> +    void *ptr, *ptr2;
>       int flags;
> -#ifdef CONFIG_DEBUG_STACK_USAGE
> -    void *ptr2;
> -#endif
>       size_t pagesz = qemu_real_host_page_size();
>   #ifdef _SC_THREAD_STACK_MIN
>       /* avoid stacks smaller than _SC_THREAD_STACK_MIN */
> @@ -619,7 +616,12 @@ void *qemu_alloc_stack(size_t *sz)
>       }
>   
>       /* Stack grows down -- guard page at the bottom. */
> -    if (mprotect(ptr, pagesz, PROT_NONE) != 0) {
> +    ptr2 = ptr;
> +#if defined(__hppa__)

Is it worth make this generic by declaring some TARGET_STACK_GROWS_UP
definition in target/foo/cpu-param.h?

> +    /* but on hppa the stack grows up, so guard the top page instead */
> +    ptr2 = ptr + *sz - pagesz;
> +#endif
> +    if (mprotect(ptr2, pagesz, PROT_NONE) != 0) {
>           perror("failed to set up stack guard page");
>           abort();
>       }