Re: [Qemu-devel] [PATCH v2 05/15] target/arm/arm-semi: Restrict use of TaskState*

[Richard Henderson <richard.henderson@linaro.org>]

On 9/16/19 7:15 AM, Peter Maydell wrote:
> The semihosting code needs accuss to the linux-user only
> TaskState pointer so it can set the semihosting errno per-thread
> for linux-user mode. At the moment we do this by having some
> ifdefs so that we define a 'ts' local in do_arm_semihosting()
> which is either a real TaskState * or just a CPUARMState *,
> depending on which mode we're compiling for.
> 
> This is awkward if we want to refactor do_arm_semihosting()
> into other functions which might need to be passed the TaskState.
> Restrict usage of the TaskState local by:
>  * making set_swi_errno() always take the CPUARMState pointer
>    and (for the linux-user version) get TaskState from that
>  * creating a new get_swi_errno() which reads the errno
>  * having the two semihosting calls which need the TaskState
>    for other purposes (SYS_GET_CMDLINE and SYS_HEAPINFO)
>    define a variable with scope restricted to just that code
> 
> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> ---
> We use 'CPUARMState *', aka 'env', rather than the other
> options of passing the ARMCPU* or the CPUState *, purely
> because it means that the later refactoring of each SYS_*
> can pass just the CPUARMState * and incidentally avoid
> an ugly ifdef caused by the implicit use of env in the
> softmmu lock_user().
> ---
>  target/arm/arm-semi.c | 111 ++++++++++++++++++++++++------------------
>  1 file changed, 63 insertions(+), 48 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~