|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [PATCH v7 49/50] x86: pass remaining stack-allocated callfunc payloads by value
Per-CPU stack mappings, to come as part of Address Space Isolation,
give some contexts page-tables which map only the stack of the CPU
they run on. A CPU in such a context cannot reach another CPU's
stack, so pointers into a caller's stack cannot be dereferenced by
call-function callees.
Convert the remaining call sites that point another CPU at a local
variable via on_selected_cpus(), with on_selected_cpus_obj_inout()
where the callee writes results back (resource_access(),
l3_cache_get()) and on_selected_cpus_obj() where it only reads
(do_write_psr_msrs(), __machine_restart()):
* XENPF_resource_op: struct resource_access on the do_platform_op()
stack; the callee writes ra.nr_done back, and ra.entries points at
a (heap) array the callee also writes - both keep working, the
former via the copy-back, the latter because the array stays
globally reachable.
* psr do_write_psr_msrs(): struct cos_write_info, which the callee
only reads; the embedded val pointer references a heap array.
* machine_restart(): the hop to the boot CPU passes delay_millisecs
with @wait == 0 and then halts this CPU forever. This is the one
asynchronous case: the halted caller's frame stays alive for the
callee (correct today), but with per-CPU mapped stacks it would no
longer be reachable. Passing by value removes the cross-CPU
reference altogether.
* XEN_SYSCTL_psr_cmt_op: struct l3_cache_info, filled by the callee.
No functional change.
Groundwork for per-CPU stack mappings, which take away a CPU's access
to other CPUs' stacks: besides get_cpu_current(), cross-CPU rendezvous
and call-function payloads are what reads another CPU's stack at run
time.
Assisted-by: Claude Code:claude-opus-4-8, Claude Code:claude-opus-5-5
Signed-off-by: George Dunlap <gwd@xxxxxxxxxxxxxx>
---
Changes in v7:
- New in this version.
---
xen/arch/x86/platform_hypercall.c | 2 +-
xen/arch/x86/psr.c | 2 +-
xen/arch/x86/shutdown.c | 4 ++--
xen/arch/x86/sysctl.c | 3 ++-
4 files changed, 6 insertions(+), 5 deletions(-)
diff --git a/xen/arch/x86/platform_hypercall.c
b/xen/arch/x86/platform_hypercall.c
index 0f3c791fc2..5516a232e4 100644
--- a/xen/arch/x86/platform_hypercall.c
+++ b/xen/arch/x86/platform_hypercall.c
@@ -876,7 +876,7 @@ ret_t do_platform_op(
if ( cpu == smp_processor_id() )
resource_access(&ra);
else
- on_selected_cpus(cpumask_of(cpu), resource_access, &ra, 1);
+ on_selected_cpus_obj_inout(cpumask_of(cpu), resource_access, &ra);
/* Copy all if succeeded or up to the failed entry. */
if ( __copy_to_guest(guest_entries, ra.entries,
diff --git a/xen/arch/x86/psr.c b/xen/arch/x86/psr.c
index cce7020868..29b8fad96e 100644
--- a/xen/arch/x86/psr.c
+++ b/xen/arch/x86/psr.c
@@ -1337,7 +1337,7 @@ static int write_psr_msrs(unsigned int socket, unsigned
int cos,
if ( cpu >= nr_cpu_ids )
return -ENOTSOCK;
- on_selected_cpus(cpumask_of(cpu), do_write_psr_msrs, &data, 1);
+ on_selected_cpus_obj(cpumask_of(cpu), do_write_psr_msrs, &data, 1);
}
return 0;
diff --git a/xen/arch/x86/shutdown.c b/xen/arch/x86/shutdown.c
index 902076cf67..820eaa7e30 100644
--- a/xen/arch/x86/shutdown.c
+++ b/xen/arch/x86/shutdown.c
@@ -550,8 +550,8 @@ void machine_restart(unsigned int delay_millisecs)
if ( get_apic_id() != boot_cpu_physical_apicid )
{
/* Send IPI to the boot CPU (logical cpu 0). */
- on_selected_cpus(cpumask_of(0), __machine_restart,
- &delay_millisecs, 0);
+ on_selected_cpus_obj(cpumask_of(0), __machine_restart,
+ &delay_millisecs, 0);
for ( ; ; )
halt();
}
diff --git a/xen/arch/x86/sysctl.c b/xen/arch/x86/sysctl.c
index 6bd4e191a7..b1a15dce97 100644
--- a/xen/arch/x86/sysctl.c
+++ b/xen/arch/x86/sysctl.c
@@ -195,7 +195,8 @@ long arch_do_sysctl(
if ( cpu == smp_processor_id() )
l3_cache_get(&info);
else
- on_selected_cpus(cpumask_of(cpu), l3_cache_get, &info, 1);
+ on_selected_cpus_obj_inout(cpumask_of(cpu), l3_cache_get,
+ &info);
ret = info.ret;
sysctl->u.psr_cmt_op.u.data = (ret ? 0 : info.size);
--
2.56.0
|
![]() |
Lists.xenproject.org is hosted with RackSpace, monitoring our |