Re: [PATCH 2/3] ipc/mach_debug.c (host_ipc_hash_info): quiet GCC warning about uninitialized variable

[Richard Braun]

On Wed, Dec 18, 2013 at 09:17:48AM +0100, Marin Ramesa wrote:
> Don't initialize to zero, rather move the initialization of size before
> the break statement. Break on the first iteration should never happen, so the
> position of initialization doesn't matter.
> 
> * ipc/mach_debug.c (host_ipc_hash_info) (size): Don't initialize to zero.
> (host_ipc_hash_info) (size): Move initialization before the break statement.
> 
> ---
>  ipc/mach_debug.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/ipc/mach_debug.c b/ipc/mach_debug.c
> index dd9057a..d7c454b 100644
> --- a/ipc/mach_debug.c
> +++ b/ipc/mach_debug.c
> @@ -112,7 +112,7 @@ host_ipc_hash_info(
>       mach_msg_type_number_t          *countp)
>  {
>       vm_offset_t addr;
> -     vm_size_t size = 0;     /* Suppress gcc warning */
> +     vm_size_t size;
>       hash_info_bucket_t *info;
>       unsigned int potential, actual;
>       kern_return_t kr;
> @@ -127,6 +127,9 @@ host_ipc_hash_info(
>  
>       for (;;) {
>               actual = ipc_hash_info(info, potential);
> +
> +             size = round_page(actual * sizeof *info);
> +
>               if (actual <= potential)
>                       break;
>  
> @@ -135,7 +138,6 @@ host_ipc_hash_info(
>               if (info != *infop)
>                       kmem_free(ipc_kernel_map, addr, size);

See that kmem_free() right here, it uses the size computed during the
previous iteration. It works as it is because ipc_hash_info returns
ipc_hash_global_size, a value that doesn't change after initialization.
But it will break if the hash table is made resizable.

> -             size = round_page(actual * sizeof *info);
>               kr = kmem_alloc_pageable(ipc_kernel_map, &addr, size);
>               if (kr != KERN_SUCCESS)
>                       return KERN_RESOURCE_SHORTAGE;

-- 
Richard Braun