Hey Kevin,
On Tue, Apr 21, 2026 at 03:42:49PM +0100, Kevin Brodsky wrote:
> Commit 2e8a1acea859 ("arm64: signal: Improve POR_EL0 handling to
> avoid uaccess failures") delayed the write to POR_EL0 in
> rt_sigreturn to avoid spurious uaccess failures. This change however
> relies on the poe_context frame record being present: on a system
> supporting POE, calling sigreturn without a poe_context record now
> results in writing arbitrary data from the kernel stack into POR_EL0.
>
> Fix this by adding a valid_fields member to struct
> user_access_state, and zeroing the struct on allocation.
> restore_poe_context() then indicates that the por_el0 field is valid
> by setting the corresponding bit in valid_fields, and
> restore_user_access_state() only touches POR_EL0 if there is a valid
> value to set it to. This is in line with how POR_EL0 was originally
> handled; all frame records are currently optional, except
> fpsimd_context.
>
> restore_user_access_state() is also called if setting up the signal
> frame fails, so we also initialise valid_fields in that case. For
> consistency, setup_sigframe() now also checks valid_fields to decide
> whether to write a poe_context record, avoiding another call to
> system_supports_poe().
>
> Fixes: 2e8a1acea859 ("arm64: signal: Improve POR_EL0 handling to avoid
> uaccess failures")
> Reported-by: Will Deacon <[email protected]>
> Signed-off-by: Kevin Brodsky <[email protected]>
Thanks for fixing this. I think your patch is correct, but I have a
couple of comments inline. Please let me know what you think.
> ---
> arch/arm64/kernel/signal.c | 19 ++++++++++++++-----
> 1 file changed, 14 insertions(+), 5 deletions(-)
>
> diff --git a/arch/arm64/kernel/signal.c b/arch/arm64/kernel/signal.c
> index 08ffc5a5aea4..3f17aed5b4f0 100644
> --- a/arch/arm64/kernel/signal.c
> +++ b/arch/arm64/kernel/signal.c
> @@ -67,6 +67,8 @@ struct rt_sigframe_user_layout {
> unsigned long end_offset;
> };
>
> +#define UA_STATE_HAS_POR_EL0 BIT(0)
> +
> /*
> * Holds any EL0-controlled state that influences unprivileged memory
> accesses.
> * This includes both accesses done in userspace and uaccess done in the
> kernel.
> @@ -74,8 +76,12 @@ struct rt_sigframe_user_layout {
> * This state needs to be carefully managed to ensure that it doesn't cause
> * uaccess to fail when setting up the signal frame, and the signal handler
> * itself also expects a well-defined state when entered.
> + *
> + * The valid_fields member is a bitfield (see UA_STATE_HAS_*), specifying
> which
> + * of the remaining fields is valid (has been set to a value).
> */
> struct user_access_state {
> + unsigned int valid_fields;
> u64 por_el0;
> };
Do you think it would be worth adding some accessors to make it easier
to keep the flags in sync? For example:
/* Stores por_el0 into uas->por_el0 and sets UA_STATE_HAS_POR_EL0 */
void set_ua_state_por_el0(struct user_access_state *uas, u64 por_el0);
/*
* If UA_STATE_HAS_POR_EL0, *por_el0 = uas->por_el0 and return 0.
* Otherwise, return -ENOENT.
*/
int get_ua_state_por_el0(struct user_access_state *uas, u64 *por_el0);
WDYT?
> @@ -1095,7 +1104,7 @@ SYSCALL_DEFINE0(rt_sigreturn)
> {
> struct pt_regs *regs = current_pt_regs();
> struct rt_sigframe __user *frame;
> - struct user_access_state ua_state;
> + struct user_access_state ua_state = {0};
nit: {} should do (no need for the '0'). Same in setup_rt_frame().
Will
