x86/hypercall: fix gnttab hypercall args conditional build on pvshim
PATCH 1/1 x86/hypercall: fix gnttab hypercall args conditional build on pvshim
A pvshim build doesn't require the grant table functionality built in, but it does require knowing the number of arguments the hypercall has so the hypercall parameter clobbering works properly.
Note this hasn't been detected by osstest because the tools pvshim build is done without debug enabled, so the hypercall parameter clobbering doesn't happen.
--- xen/arch/x86/hypercall.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
xen/arch/x86/hypercall.c +1 -1
@@ -47,7 +47,7 @@ const hypercall_args_t hypercall_args_table[NR_hypercalls] = ARGS(xen_version, 2), ARGS(console_io, 3), ARGS(physdev_op_compat, 1), -#ifdef CONFIG_GRANT_TABLE +#if defined(CONFIG_GRANT_TABLE) || defined(CONFIG_PV_SHIM) ARGS(grant_table_op, 3),#endif ARGS(vm_assist, 2),
Signature · 1 line
2.29.2
xen/arch/x86/hypercall.c const hypercall_args_t hypercall_args_table[NR_hypercalls] = · line 53
Quoted 11 lines · Roger Pau Monne
xen/arch/x86/hypercall.c +1
-1@@ -47,7 +47,7 @@ const hypercall_args_t hypercall_args_table[NR_hypercalls] = ARGS(xen_version, 2), ARGS(console_io, 3), ARGS(physdev_op_compat, 1), -#ifdef CONFIG_GRANT_TABLE +#if defined(CONFIG_GRANT_TABLE) || defined(CONFIG_PV_SHIM) ARGS(grant_table_op, 3),#endif ARGS(vm_assist, 2),
This is correct when a shim-enabled build runs as shim, but not when it runs as normal hypervisor. Just like the hypercall handler gets patched into the hypercall table (in pv_shim_setup_dom()), the argument count will also want updating there, I think.
Jan
Quoted 10 lines · Jan Beulich
Quoted 9 lines · Roger Pau Monne
xen/arch/x86/hypercall.c +1
-1@@ -47,7 +47,7 @@ const hypercall_args_t hypercall_args_table[NR_hypercalls] = ARGS(xen_version, 2), ARGS(console_io, 3), ARGS(physdev_op_compat, 1), -#ifdef CONFIG_GRANT_TABLE +#if defined(CONFIG_GRANT_TABLE ) || defined(CONFIG_PV_SHIM ) ARGS(grant_table_op, 3),
Quoted 8 lines
xen/arch/x86/hypercall.c const hypercall_args_t hypercall_args_table[NR_hypercalls] = · line 53
Quoted 2 lines
xen/arch/x86/hypercall.c +0
-0#endif ARGS(vm_assist, 2),This is correct when a shim-enabled build runs as shim, but not when it runs as normal hypervisor. Just like the hypercall handler gets patched into the hypercall table (in
pv_shim_setup_dom()), the argument count will also want updating there, I think.
Having the argument count set when the hypercall handler is NULL is fine, as Xen won't get into processing hypercall_args_table if the handler is NULL. While it's true that can be fixed at run time in the same way that we patch the handler it seems more cumbersome, that's why I went for this IMO simpler fix.
Thanks, Roger.
Jan Beulich
Reviewed-byQuoted 20 lines · Roger Pau Monné
Quoted 10 lines · Jan Beulich
Quoted 9 lines · Roger Pau Monne
xen/arch/x86/hypercall.c +1
-1@@ -47,7 +47,7 @@ const hypercall_args_t hypercall_args_table[NR_hypercalls] = ARGS(xen_version, 2), ARGS(console_io, 3), ARGS(physdev_op_compat, 1), -#ifdef CONFIG_GRANT_TABLE +#if defined(CONFIG_GRANT_TABLE ) || defined(CONFIG_PV_SHIM ) ARGS(grant_table_op, 3),Quoted 8 lines
xen/arch/x86/hypercall.c const hypercall_args_t hypercall_args_table[NR_hypercalls] = · line 53
Quoted 2 lines
xen/arch/x86/hypercall.c +0
-0#endif ARGS(vm_assist, 2),This is correct when a shim-enabled build runs as shim, but not when it runs as normal hypervisor. Just like the hypercall handler gets patched into the hypercall table (in
pv_shim_setup_dom()), the argument count will also want updating there, I think.
Quoted 3 lines
Having the argument count set when the hypercall handler is NULL is fine, as Xen won't get into processing
hypercall_args_tableif the handler is NULL.
Oh, good point.
Quoted 3 lines
While it's true that can be fixed at run time in the same way that we patch the handler it seems more cumbersome, that's why I went for this IMO simpler fix.
Agreed:
Jan
Quoted 21 lines · Jan Beulich
Quoted 20 lines · Roger Pau Monné
Quoted 18 lines · Jan Beulich
xen/arch/x86/hypercall.c const hypercall_args_t hypercall_args_table[NR_hypercalls] = · line 53
Quoted 11 lines · Roger Pau Monne
xen/arch/x86/hypercall.c +1
-1@@ -47,7 +47,7 @@ const hypercall_args_t hypercall_args_table[NR_hypercalls] = ARGS(xen_version, 2), ARGS(console_io, 3), ARGS(physdev_op_compat, 1), -#ifdef CONFIG_GRANT_TABLE +#if defined(CONFIG_GRANT_TABLE) || defined(CONFIG_PV_SHIM) ARGS(grant_table_op, 3),#endif ARGS(vm_assist, 2),This is correct when a shim-enabled build runs as shim, but not when it runs as normal hypervisor. Just like the hypercall handler gets patched into the hypercall table (in
pv_shim_setup_dom()), the argument count will also want updating there, I think.
Quoted 5 lines
Quoted 3 lines
Having the argument count set when the hypercall handler is NULL is fine, as Xen won't get into processing
hypercall_args_tableif the handler is NULL.Oh, good point.
Albeit then - why not drop the #ifdef altogether?
Jan
Quoted 25 lines · Jan Beulich
Quoted 20 lines · Roger Pau Monné
Quoted 10 lines · Jan Beulich
Quoted 9 lines · Roger Pau Monne
xen/arch/x86/hypercall.c +1
-1@@ -47,7 +47,7 @@ const hypercall_args_t hypercall_args_table[NR_hypercalls] = ARGS(xen_version, 2), ARGS(console_io, 3), ARGS(physdev_op_compat, 1), -#ifdef CONFIG_GRANT_TABLE +#if defined(CONFIG_GRANT_TABLE ) || defined(CONFIG_PV_SHIM ) ARGS(grant_table_op, 3),Quoted 8 lines
xen/arch/x86/hypercall.c const hypercall_args_t hypercall_args_table[NR_hypercalls] = · line 53
Quoted 2 lines
xen/arch/x86/hypercall.c +0
-0#endif ARGS(vm_assist, 2),This is correct when a shim-enabled build runs as shim, but not when it runs as normal hypervisor. Just like the hypercall handler gets patched into the hypercall table (in
pv_shim_setup_dom()), the argument count will also want updating there, I think.Quoted 3 lines
Having the argument count set when the hypercall handler is NULL is fine, as Xen won't get into processing
hypercall_args_tableif the handler is NULL.
Quoted 8 lines
Oh, good point.
Quoted 3 lines
While it's true that can be fixed at run time in the same way that we patch the handler it seems more cumbersome, that's why I went for this IMO simpler fix.
Agreed:
Reviewed-by Jan Beulich <[email protected]>
Thanks.
In fact I'm not sure it's helpful to have any build time conditionals in hypercall_args_table, but let's leave that for another fix.
Roger.