On 13/02/2023 11:30, Tudor Cretu wrote:
On 10-02-2023 09:16, Kevin Brodsky wrote:
sp_top is only used to compute the size to provide to access_ok(). It should therefore be just an address, not a capability (in PCuABI). GCC rightly warns about this situation, as subtracting two capabilities is ambiguous (it may be interpreted as a capability operation subtracting the address of the LHS with the address of the RHS, while here we expect a simple integer subtraction).
Fixes: ("linux/sched/signal.h: Modify the stack pointer to user_uintptr_t") Signed-off-by: Kevin Brodsky kevin.brodsky@arm.com
arch/arm64/kernel/signal.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/kernel/signal.c b/arch/arm64/kernel/signal.c index b804457a36b2..80289e90fc66 100644 --- a/arch/arm64/kernel/signal.c +++ b/arch/arm64/kernel/signal.c
nit: I'm not sure on this, should we include linux/types.h explicitly for ptraddr_t?
We could, but it's not essential, as we already include <linux/kernel.h> and it itself includes <linux/types.h>, which is something that I think we can reasonably rely on.
Kevin
Thanks, Tudor
@@ -1001,7 +1001,8 @@ static int setup_sigframe(struct rt_sigframe_user_layout *user, static int get_sigframe(struct rt_sigframe_user_layout *user, struct ksignal *ksig, struct pt_regs *regs) { - user_uintptr_t sp, sp_top; + user_uintptr_t sp; + ptraddr_t sp_top; int err; init_user_layout(user); @@ -1009,7 +1010,8 @@ static int get_sigframe(struct rt_sigframe_user_layout *user, if (err) return err; - sp = sp_top = sigsp(signal_sp(regs), ksig); + sp = sigsp(signal_sp(regs), ksig); + sp_top = (ptraddr_t)sp; sp = round_down(sp - sizeof(struct frame_record), 16); user->next_frame = (struct frame_record __user *)sp;