Re: [Qemu-devel] [PATCH 4/6] linux-user: Provide safe_syscall for aarch64

[Peter Maydell]

On 13 June 2016 at 22:45, Richard Henderson <address@hidden> wrote:
> Signed-off-by: Richard Henderson <address@hidden>
> ---
>  linux-user/host/aarch64/hostdep.h          | 34 ++++++++++++++
>  linux-user/host/aarch64/safe-syscall.inc.S | 72 
> ++++++++++++++++++++++++++++++
>  2 files changed, 106 insertions(+)
>  create mode 100644 linux-user/host/aarch64/hostdep.h
>  create mode 100644 linux-user/host/aarch64/safe-syscall.inc.S
>
> diff --git a/linux-user/host/aarch64/hostdep.h 
> b/linux-user/host/aarch64/hostdep.h
> new file mode 100644
> index 0000000..0ff7985
> --- /dev/null
> +++ b/linux-user/host/aarch64/hostdep.h
> @@ -0,0 +1,34 @@
> +/*
> + * hostdep.h : things which are dependent on the host architecture
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + */
> +
> +#ifndef QEMU_HOSTDEP_H
> +#define QEMU_HOSTDEP_H
> +
> +/* We have a safe-syscall.inc.S */
> +#define HAVE_SAFE_SYSCALL
> +
> +#ifndef __ASSEMBLER__
> +
> +/* These are defined by the safe-syscall.inc.S file */
> +extern char safe_syscall_start[];
> +extern char safe_syscall_end[];
> +
> +/* Adjust the signal context to rewind out of safe-syscall if we're in it */
> +static inline void rewind_if_in_safe_syscall(void *puc)
> +{
> +    struct ucontext *uc = puc;
> +    __u64 *pcreg = &uc->uc_mcontext.pc;
> +
> +    if (*pcreg > (uintptr_t)safe_syscall_start
> +        && *pcreg < (uintptr_t)safe_syscall_end) {
> +        *pcreg = (uintptr_t)safe_syscall_start;
> +    }
> +}
> +
> +#endif /* __ASSEMBLER__ */
> +
> +#endif
> diff --git a/linux-user/host/aarch64/safe-syscall.inc.S 
> b/linux-user/host/aarch64/safe-syscall.inc.S
> new file mode 100644
> index 0000000..5416b90
> --- /dev/null
> +++ b/linux-user/host/aarch64/safe-syscall.inc.S
> @@ -0,0 +1,72 @@
> +/*
> + * safe-syscall.inc.S : host-specific assembly fragment
> + * to handle signals occurring at the same time as system calls.
> + * This is intended to be included by linux-user/safe-syscall.S
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
> + * See the COPYING file in the top-level directory.
> + */
> +
> +        .global safe_syscall_base
> +        .global safe_syscall_start
> +        .global safe_syscall_end
> +        .type   safe_syscall_base, #function
> +        .type   safe_syscall_start, #function
> +        .type   safe_syscall_end, #function

_start and _end aren't function entry points, so is it OK
to mark them as functions?

(The 'as' manual doesn't document what setting .type does in much
detail...)

> +
> +        /* This is the entry point for making a system call. The calling
> +         * convention here is that of a C varargs function with the
> +         * first argument an 'int *' to the signal_pending flag, the
> +         * second one the system call number (as a 'long'), and all further
> +         * arguments being syscall arguments (also 'long').
> +         * We return a long which is the syscall's return value, which
> +         * may be negative-errno on failure. Conversion to the
> +         * -1-and-errno-set convention is done by the calling wrapper.
> +         */
> +safe_syscall_base:
> +        .cfi_startproc
> +        /* The syscall calling convention isn't the same as the
> +         * C one:
> +         * we enter with x0 == *signal_pending
> +         *               x1 == syscall number
> +         *               x2 ... x7, (stack) == syscall arguments
> +         *               and return the result in x0
> +         * and the syscall instruction needs
> +         *               x8 == syscall number
> +         *               x0 ... x6 == syscall arguments
> +         *               and returns the result in x0
> +         * Shuffle everything around appropriately.
> +         */

http://man7.org/linux/man-pages/man2/syscall.2.html
doesn't mention a syscall argument in x7, just x0..x5.

> +       mov     x9, x0          /* signal_pending pointer */
> +       mov     w8, w1          /* syscall number */

Seems a bit odd to use a 32-bit move for this when our input
calling convention has it as 64 bits and the kernel's calling
convention has it as 64 bits.

> +       mov     x0, x2          /* syscall arguments */
> +       mov     x1, x3
> +       mov     x2, x4
> +       mov     x3, x5
> +       mov     x4, x6
> +       mov     x6, x7
> +       ldr     x7, [sp]
> +
> +        /* This next sequence of code works in conjunction with the
> +         * rewind_if_safe_syscall_function(). If a signal is taken
> +         * and the interrupted PC is anywhere between 'safe_syscall_start'
> +         * and 'safe_syscall_end' then we rewind it to 'safe_syscall_start'.
> +         * The code sequence must therefore be able to cope with this, and
> +         * the syscall instruction must be the final one in the sequence.
> +         */
> +safe_syscall_start:
> +        /* if signal_pending is non-zero, don't do the call */
> +       ldr     w10, [x9]
> +       cbnz    w10, 0f
> +        svc    0x0
> +safe_syscall_end:
> +        /* code path for having successfully executed the syscall */
> +        ret
> +
> +0:
> +        /* code path when we didn't execute the syscall */
> +        mov     x0, #-TARGET_ERESTARTSYS
> +        ret
> +        .cfi_endproc
> +
> +        .size   safe_syscall_base, .-safe_syscall_base
> --
> 2.5.5
>

thanks
-- PMM