EFI x86: pass firmware call parameters on the stack
PATCH 1/1 EFI x86: pass firmware call parameters on the stack
When calling into an EFI firmware, the parameters need to be passed on the stack. The recent change to use -mregparm=3 breaks x86 EFI support. This patch is needed to allow the new Intel-based Macs to suspend to ram when run in EFI mode (efi.get_time is called during the suspend phase).
---
[As I couldn't find an official maintainer for the linux/efi.h file and the file header is quite old, I'm Cc:ing the last 2 commiters.]
This patch fixes the issue for x86, but the file is also used by IA64. I would have used asmlinkage to force arguments on the stack, but it has a special meaning on IA64, thus I used a raw regparm(0) GCC attribute. This attribute is documented only for x86, I hope it has no side effect on other archs.
include/linux/efi.h +20 -10
@@ -157,25 +157,35 @@ typedef struct {unsigned long reset_system; } efi_runtime_services_t; -typedef efi_status_t efi_get_time_t (efi_time_t *tm,efi_time_cap_t *tc); -typedef efi_status_t efi_set_time_t (efi_time_t *tm); +typedef efi_status_t efi_get_time_t (efi_time_t *tm,efi_time_cap_t *tc) + __attribute__((regparm (0 ))); +typedef efi_status_t efi_set_time_t (efi_time_t *tm) + __attribute__((regparm (0 )));typedef efi_status_t efi_get_wakeup_time_t (efi_bool_t *enabled,efi_bool_t *pending, -efi_time_t *tm); -typedef efi_status_t efi_set_wakeup_time_t (efi_bool_t enabled,efi_time_t *tm); +efi_time_t *tm) + __attribute__((regparm (0 ))); +typedef efi_status_t efi_set_wakeup_time_t (efi_bool_t enabled,efi_time_t *tm) + __attribute__((regparm (0 )));typedef efi_status_t efi_get_variable_t (efi_char16_t *name,efi_guid_t *vendor,u32 *attr, -unsigned long *data_size,void *data); +unsigned long *data_size,void *data) + __attribute__((regparm (0 )));typedef efi_status_t efi_get_next_variable_t (unsigned long *name_size,efi_char16_t *name, -efi_guid_t *vendor); +efi_guid_t *vendor) + __attribute__((regparm (0 )));typedef efi_status_t efi_set_variable_t (efi_char16_t *name,efi_guid_t *vendor,unsigned long attr,unsigned long data_size, -void *data); -typedef efi_status_t efi_get_next_high_mono_count_t (u32 *count); +void *data) + __attribute__((regparm (0 ))); +typedef efi_status_t efi_get_next_high_mono_count_t (u32 *count) + __attribute__((regparm (0 )));typedef void efi_reset_system_t (int reset_type,efi_status_t status, -unsigned long data_size,efi_char16_t *data); +unsigned long data_size,efi_char16_t *data) + __attribute__((regparm (0 )));typedef efi_status_t efi_set_virtual_address_map_t (unsigned long memory_map_size,unsigned long descriptor_size,u32 descriptor_version, -efi_memory_desc_t *virtual_map); +efi_memory_desc_t *virtual_map) + __attribute__((regparm (0 )));/* * EFI Configuration Table and GUID definitions
Quoted 6 lines · On Tue, 30 Jan 2007 20:01:18 +0100 Frédéric Riss <[email protected]>…
When calling into an EFI firmware, the parameters need to be passed on the stack. The recent change to use -mregparm=3 breaks x86 EFI support. This patch is needed to allow the new Intel-based Macs to suspend to ram when run in EFI mode (efi.get_time is called during the suspend phase).
Quoted 6 lines
Signed-off-by Frederic Riss <[email protected]>---
[As I couldn't find an official maintainer for the
linux/efi.hfile and the file header is quite old, I'm Cc:ing the last 2 commiters.]
Matt normally looks after EFI.
Quoted 5 lines
This patch fixes the issue for x86, but the file is also used by IA64. I would have used asmlinkage to force arguments on the stack, but it has a special meaning on IA64, thus I used a raw regparm(0) GCC attribute. This attribute is documented only for x86, I hope it has no side effect on other archs.
hm, this sounds like a fairly serious problem. Has this been runtime tested on ia64 and x86_64>
Quoted 8 lines · Andrew Morton
Quoted 5 lines
This patch fixes the issue for x86, but the file is also used by IA64. I would have used asmlinkage to force arguments on the stack, but it has a special meaning on IA64, thus I used a raw regparm(0) GCC attribute. This attribute is documented only for x86, I hope it has no side effect on other archs.
hm, this sounds like a fairly serious problem. Has this been runtime tested on ia64 and
x86_64>
Not by me, the only EFI box I've access to is my Mac Mini.
Fred.
Quoted 10 lines · Frédéric Riss
When calling into an EFI firmware, the parameters need to be passed on the stack. The recent change to use -mregparm=3 breaks x86 EFI support. This patch is needed to allow the new Intel-based Macs to suspend to ram when run in EFI mode (efi.get_time is called during the suspend phase).
...
This patch fixes the issue for x86, but the file is also used by IA64. I would have used asmlinkage to force arguments on the stack, but it has a special meaning on IA64, thus I used a raw regparm(0) GCC attribute. This attribute is documented only for x86, I hope it has no side effect on other archs.
It seems wrong to put this sort of architecture-dependent stuff in an architecture-independent file. If EFI needs a specific calling convention on x86, x86 should provide wrappers that guarantee that convention. ia64 already provides such wrappers (phys_get_time(), virt_get_time(), etc).
Quoted 53 lines
include/linux/efi.h +20
-10@@ -157,25 +157,35 @@ typedef struct {unsigned long reset_system; } efi_runtime_services_t; -typedef efi_status_t efi_get_time_t (efi_time_t *tm,efi_time_cap_t *tc); -typedef efi_status_t efi_set_time_t (efi_time_t *tm); +typedef efi_status_t efi_get_time_t (efi_time_t *tm,efi_time_cap_t *tc) + __attribute__((regparm (0 ))); +typedef efi_status_t efi_set_time_t (efi_time_t *tm) + __attribute__((regparm (0 )));typedef efi_status_t efi_get_wakeup_time_t (efi_bool_t *enabled,efi_bool_t *pending, -efi_time_t *tm); -typedef efi_status_t efi_set_wakeup_time_t (efi_bool_t enabled,efi_time_t *tm); +efi_time_t *tm) + __attribute__((regparm (0 ))); +typedef efi_status_t efi_set_wakeup_time_t (efi_bool_t enabled,efi_time_t *tm) + __attribute__((regparm (0 )));typedef efi_status_t efi_get_variable_t (efi_char16_t *name,efi_guid_t *vendor,u32 *attr, -unsigned long *data_size,void *data); +unsigned long *data_size,void *data) + __attribute__((regparm (0 )));typedef efi_status_t efi_get_next_variable_t (unsigned long *name_size,efi_char16_t *name, -efi_guid_t *vendor); +efi_guid_t *vendor) + __attribute__((regparm (0 )));typedef efi_status_t efi_set_variable_t (efi_char16_t *name,efi_guid_t *vendor,unsigned long attr,unsigned long data_size, -void *data); -typedef efi_status_t efi_get_next_high_mono_count_t (u32 *count); +void *data) + __attribute__((regparm (0 ))); +typedef efi_status_t efi_get_next_high_mono_count_t (u32 *count) + __attribute__((regparm (0 )));typedef void efi_reset_system_t (int reset_type,efi_status_t status, -unsigned long data_size,efi_char16_t *data); +unsigned long data_size,efi_char16_t *data) + __attribute__((regparm (0 )));typedef efi_status_t efi_set_virtual_address_map_t (unsigned long memory_map_size,unsigned long descriptor_size,u32 descriptor_version, -efi_memory_desc_t *virtual_map); +efi_memory_desc_t *virtual_map) + __attribute__((regparm (0 )));/* * EFI Configuration Table and GUID definitions
Quoted 12 lines · Bjorn Helgaas
Quoted 10 lines · Frédéric Riss
When calling into an EFI firmware, the parameters need to be passed on the stack. The recent change to use -mregparm=3 breaks x86 EFI support. This patch is needed to allow the new Intel-based Macs to suspend to ram when run in EFI mode (efi.get_time is called during the suspend phase).
...
This patch fixes the issue for x86, but the file is also used by IA64. I would have used asmlinkage to force arguments on the stack, but it has a special meaning on IA64, thus I used a raw regparm(0) GCC attribute. This attribute is documented only for x86, I hope it has no side effect on other archs.
Quoted 5 lines
It seems wrong to put this sort of architecture-dependent stuff in an architecture-independent file. If EFI needs a specific calling convention on x86, x86 should provide wrappers that guarantee that convention. ia64 already provides such wrappers (
phys_get_time(),virt_get_time(), etc).
[ Disclaimer: I don't know much about EFI. I wanted to download the spec, but it required me to sign some sort of agreement about needing a license to write an implementation... I ran away. ]
You're totally right, I first thought it wasn't possible, but the functions in the 'struct efi' can be set to whatever is wanted. Bellow's a patch which won't have any impact on ia64 and fixes the issue on my Mac Mini. I don't think x86_64 can be configured to boot from EFI, so this shouldn't cause any regression on that front. How does that look?
When calling into the EFI firmware, the parameters need to be passed on the stack. The recent change to use -mregparm=3 breaks x86 EFI support. This patch is needed to allow the new Intel-based Macs to suspend to ram (efi.get_time is called during the suspend phase).
---
arch/i386/kernel/efi.c +73 -16
@@ -472,6 +472,70 @@ static inline void __init check_range_fo } } +/* +* Wrap all the virtual calls in a way that forces the parameters on the stack. +*/ + +#define efi_call_virt(f, args...) \ + ((efi_##f##_t __attribute__((regparm(0)))*)efi.systab->runtime->f)(args) + +static efi_status_t virt_efi_get_time (efi_time_t *tm,efi_time_cap_t *tc) +{ +return efi_call_virt (get_time, tm, tc); +} + +static efi_status_t virt_efi_set_time (efi_time_t *tm) +{ +return efi_call_virt (set_time, tm); +} + +static efi_status_t virt_efi_get_wakeup_time (efi_bool_t *enabled, +efi_bool_t *pending, +efi_time_t *tm) +{ +return efi_call_virt (get_wakeup_time, enabled, pending, tm); +} + +static efi_status_t virt_efi_set_wakeup_time (efi_bool_t enabled, +efi_time_t *tm) +{ +return efi_call_virt (set_wakeup_time, enabled, tm); +} + +static efi_status_t virt_efi_get_variable (efi_char16_t *name, +efi_guid_t *vendor,u32 *attr, +unsigned long *data_size,void *data) +{ +return efi_call_virt (get_variable, name, vendor, attr, data_size, data); +} + +static efi_status_t virt_efi_get_next_variable (unsigned long *name_size, +efi_char16_t *name, +efi_guid_t *vendor) +{ +return efi_call_virt (get_next_variable, name_size, name, vendor); +} + +static efi_status_t virt_efi_set_variable (efi_char16_t *name, +efi_guid_t *vendor, +unsigned long attr, +unsigned long data_size,void *data) +{ +return efi_call_virt (set_variable, name, vendor, attr, data_size, data); +} + +static efi_status_t virt_efi_get_next_high_mono_count (u32 *count) +{ +return efi_call_virt (get_next_high_mono_count, count); +} + +static void virt_efi_reset_system (int reset_type,efi_status_t status, +unsigned long data_size, +efi_char16_t *data) +{ +efi_call_virt (reset_system, reset_type, status, data_size, data); +} +/* * This function will switch the EFI runtime services to virtual mode. * Essentially, look through the EFI memmap and map every region that
@@ -525,22 +589,15 @@ void __init efi_enter_virtual_mode(void)* pointers in the runtime service table to the new virtual addresses. */ - efi.get_time = (efi_get_time_t *) efi.systab->runtime->get_time; - efi.set_time = (efi_set_time_t *) efi.systab->runtime->set_time; - efi.get_wakeup_time = (efi_get_wakeup_time_t *) - efi.systab->runtime->get_wakeup_time; - efi.set_wakeup_time = (efi_set_wakeup_time_t *) - efi.systab->runtime->set_wakeup_time; - efi.get_variable = (efi_get_variable_t *) - efi.systab->runtime->get_variable; - efi.get_next_variable = (efi_get_next_variable_t *) - efi.systab->runtime->get_next_variable; - efi.set_variable = (efi_set_variable_t *) - efi.systab->runtime->set_variable; - efi.get_next_high_mono_count = (efi_get_next_high_mono_count_t *) - efi.systab->runtime->get_next_high_mono_count; - efi.reset_system = (efi_reset_system_t *) - efi.systab->runtime->reset_system; + efi.get_time = virt_efi_get_time; + efi.set_time = virt_efi_set_time; + efi.get_wakeup_time = virt_efi_get_wakeup_time; + efi.set_wakeup_time = virt_efi_set_wakeup_time; + efi.get_variable = virt_efi_get_variable; + efi.get_next_variable = virt_efi_get_next_variable; + efi.set_variable = virt_efi_set_variable; + efi.get_next_high_mono_count = virt_efi_get_next_high_mono_count; + efi.reset_system = virt_efi_reset_system; }void __init
Quoted 22 lines · Frédéric Riss
Le mardi 30 janvier 2007 à 12:17 -0700, Bjorn Helgaas a écrit :
Quoted 11 lines · Frédéric Riss
Quoted 10 lines
When calling into an EFI firmware, the parameters need to be passed on the stack. The recent change to use -mregparm=3 breaks x86 EFI support. This patch is needed to allow the new Intel-based Macs to suspend to ram when run in EFI mode (efi.get_time is called during the suspend phase).
...
This patch fixes the issue for x86, but the file is also used by IA64. I would have used asmlinkage to force arguments on the stack, but it has a special meaning on IA64, thus I used a raw regparm(0) GCC attribute. This attribute is documented only for x86, I hope it has no side effect on other archs.Quoted 5 lines
It seems wrong to put this sort of architecture-dependent stuff in an architecture-independent file. If EFI needs a specific calling convention on x86, x86 should provide wrappers that guarantee that convention. ia64 already provides such wrappers (
phys_get_time(),virt_get_time(), etc).[ Disclaimer: I don't know much about EFI. I wanted to download the spec, but it required me to sign some sort of agreement about needing a license to write an implementation... I ran away. ]
Quoted 5 lines
You're totally right, I first thought it wasn't possible, but the functions in the 'struct efi' can be set to whatever is wanted. Bellow's a patch which won't have any impact on ia64 and fixes the issue on my Mac Mini. I don't think
x86_64can be configured to boot from EFI, so this shouldn't cause any regression on that front. How does that look?
currently x86_64 kernel does not support efi, efi convention comply to MS convention. On ia32 parameter is passed on stack, on x86_64 parameter is passed by registers but that is different from x86_64 linux convention.
How about adding EFIAPI prefix before efi runtime service function, this prefix has different definition in different architecture.
thanks
bibo,mao
include/linux/efi.h renamed from include/linux/efi.h.bak +0 -0
@@ -157,22 +157,28 @@ typedef struct {
unsigned long reset_system; } efi_runtime_services_t;
-typedef efi_status_t efi_get_time_t (efi_time_t *tm, efi_time_cap_t *tc); -typedef efi_status_t efi_set_time_t (efi_time_t *tm);
-typedef efi_status_t efi_get_wakeup_time_t (efi_bool_t *enabled, efi_bool_t *pending,
+#ifdef CONFIG_X86_32 +#define EFIAPI asmlinkage
+# else
+#define EFIAPI
+# endif
+
+typedef EFIAPI efi_status_t efi_get_time_t (efi_time_t *tm, efi_time_cap_t *tc); +typedef EFIAPI efi_status_t efi_set_time_t (efi_time_t *tm);
+typedef EFIAPI efi_status_t efi_get_wakeup_time_t (efi_bool_t *enabled, efi_bool_t *pending,
efi_time_t *tm);
-typedef efi_status_t efi_set_wakeup_time_t (efi_bool_t enabled, efi_time_t *tm); -typedef efi_status_t efi_get_variable_t (efi_char16_t *name, efi_guid_t *vendor, u32 *attr, +typedef EFIAPI efi_status_t efi_set_wakeup_time_t (efi_bool_t enabled, efi_time_t *tm); +typedef EFIAPI efi_status_t efi_get_variable_t (efi_char16_t *name, efi_guid_t *vendor, u32 *attr,
unsigned long *data_size, void *data); -typedef efi_status_t efi_get_next_variable_t (unsigned long *name_size, efi_char16_t *name, +typedef EFIAPI efi_status_t efi_get_next_variable_t (unsigned long *name_size, efi_char16_t *name, efi_guid_t *vendor); -typedef efi_status_t efi_set_variable_t (efi_char16_t *name, efi_guid_t *vendor, +typedef EFIAPI efi_status_t efi_set_variable_t (efi_char16_t *name, efi_guid_t *vendor, unsigned long attr, unsigned long data_size, void *data);
-typedef efi_status_t efi_get_next_high_mono_count_t (u32 *count); -typedef void efi_reset_system_t (int reset_type, efi_status_t status, +typedef EFIAPI efi_status_t efi_get_next_high_mono_count_t (u32 *count); +typedef EFIAPI void efi_reset_system_t (int reset_type, efi_status_t status,
unsigned long data_size, efi_char16_t *data); -typedef efi_status_t efi_set_virtual_address_map_t (unsigned long memory_map_size, +typedef EFIAPI efi_status_t efi_set_virtual_address_map_t (unsigned long memory_map_size, unsigned long descriptor_size, u32 descriptor_version, efi_memory_desc_t *virtual_map);
Quoted 126 lines
When calling into the EFI firmware, the parameters need to be passed on the stack. The recent change to use -mregparm=3 breaks x86 EFI support. This patch is needed to allow the new Intel-based Macs to suspend to ram (efi.get_time is called during the suspend phase).
Signed-off-by Frederic Riss <[email protected]>---
arch/i386/kernel/efi.c +73
-17@@ -472,6 +472,70 @@ static inline void __init check_range_fo } } +/* +* Wrap all the virtual calls in a way that forces the parameters on the stack. +*/ + +#define efi_call_virt(f, args...) \ + ((efi_##f##_t __attribute__((regparm(0)))*)efi.systab->runtime->f)(args) + +static efi_status_t virt_efi_get_time (efi_time_t *tm,efi_time_cap_t *tc) +{ +return efi_call_virt (get_time, tm, tc); +} + +static efi_status_t virt_efi_set_time (efi_time_t *tm) +{ +return efi_call_virt (set_time, tm); +} + +static efi_status_t virt_efi_get_wakeup_time (efi_bool_t *enabled, +efi_bool_t *pending, +efi_time_t *tm) +{ +return efi_call_virt (get_wakeup_time, enabled, pending, tm); +} + +static efi_status_t virt_efi_set_wakeup_time (efi_bool_t enabled, +efi_time_t *tm) +{ +return efi_call_virt (set_wakeup_time, enabled, tm); +} + +static efi_status_t virt_efi_get_variable (efi_char16_t *name, +efi_guid_t *vendor,u32 *attr, +unsigned long *data_size,void *data) +{ +return efi_call_virt (get_variable, name, vendor, attr, data_size, data); +} + +static efi_status_t virt_efi_get_next_variable (unsigned long *name_size, +efi_char16_t *name, +efi_guid_t *vendor) +{ +return efi_call_virt (get_next_variable, name_size, name, vendor); +} + +static efi_status_t virt_efi_set_variable (efi_char16_t *name, +efi_guid_t *vendor, +unsigned long attr, +unsigned long data_size,void *data) +{ +return efi_call_virt (set_variable, name, vendor, attr, data_size, data); +} + +static efi_status_t virt_efi_get_next_high_mono_count (u32 *count) +{ +return efi_call_virt (get_next_high_mono_count, count); +} + +static void virt_efi_reset_system (int reset_type,efi_status_t status, +unsigned long data_size, +efi_char16_t *data) +{ +efi_call_virt (reset_system, reset_type, status, data_size, data); +} +/* * This function will switch the EFI runtime services to virtual mode. * Essentially, look through the EFI memmap and map every region that @@ -525,22 +589,15 @@ void __init efi_enter_virtual_mode(void)* pointers in the runtime service table to the new virtual addresses. */ - efi.get_time = (efi_get_time_t *) efi.systab->runtime->get_time; - efi.set_time = (efi_set_time_t *) efi.systab->runtime->set_time; - efi.get_wakeup_time = (efi_get_wakeup_time_t *) - efi.systab->runtime->get_wakeup_time; - efi.set_wakeup_time = (efi_set_wakeup_time_t *) - efi.systab->runtime->set_wakeup_time; - efi.get_variable = (efi_get_variable_t *) - efi.systab->runtime->get_variable; - efi.get_next_variable = (efi_get_next_variable_t *) - efi.systab->runtime->get_next_variable; - efi.set_variable = (efi_set_variable_t *) - efi.systab->runtime->set_variable; - efi.get_next_high_mono_count = (efi_get_next_high_mono_count_t *) - efi.systab->runtime->get_next_high_mono_count; - efi.reset_system = (efi_reset_system_t *) - efi.systab->runtime->reset_system; + efi.get_time = virt_efi_get_time; + efi.set_time = virt_efi_set_time; + efi.get_wakeup_time = virt_efi_get_wakeup_time; + efi.set_wakeup_time = virt_efi_set_wakeup_time; + efi.get_variable = virt_efi_get_variable; + efi.get_next_variable = virt_efi_get_next_variable; + efi.set_variable = virt_efi_set_variable; + efi.get_next_high_mono_count = virt_efi_get_next_high_mono_count; + efi.reset_system = virt_efi_reset_system; }void __init -To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
2007/2/1, bibo,mao <[email protected]>:
Quoted 3 lines
currently
x86_64kernel does not support efi, efi convention comply to MS convention. On ia32 parameter is passed on stack, onx86_64parameter is passed by registers but that is different fromx86_64linux convention.
Is an x86_64 EFI firmware required to have 2 entry points, one x86_64 and one in ia32 mode ? If that's not the case, it means that an i386 kernel won't be able to work correctly on an EFI powered x86_64 box...
Quoted 2 lines
How about adding EFIAPI prefix before efi runtime service function, this prefix has different definition in different architecture.
Have you any objection to the second patch I posted? It's not only a matter of passing arguments, ia64 does quite a lot of things in its wrapper functions. I can easily imagine that such additional work could be added in the i386 wrappers if EFI becomes more widely used.
Quoted 9 lines · Frederic Riss
2007/2/1, bibo,mao <[email protected]>:
Quoted 3 lines
currently
x86_64kernel does not support efi, efi convention comply to MS convention. On ia32 parameter is passed on stack, onx86_64parameter is passed by registers but that is different fromx86_64linux convention.Is an
x86_64EFI firmware required to have 2 entry points, onex86_64and one in ia32 mode ? If that's not the case, it means that an i386 kernel won't be able to work correctly on an EFI poweredx86_64box...
AFAIK x86_64 EFI firmware now has only one entry point, i386 kernel can not boot up with EFI powered x86_64 box.
Quoted 8 lines
Quoted 2 lines
How about adding EFIAPI prefix before efi runtime service function, this prefix has different definition in different architecture.
Have you any objection to the second patch I posted? It's not only a matter of passing arguments, ia64 does quite a lot of things in its wrapper functions. I can easily imagine that such additional work could be added in the i386 wrappers if EFI becomes more widely used.
I do not object your second patch if there will be more wrapper functions to be added in i386 efi code. I have to give up my patch:)
thanks
bibo,mao