Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
/*-
|
|
|
|
* Copyright (c) 2004 Marcel Moolenaar
|
|
|
|
* Copyright (c) 2001 Doug Rabson
|
2018-09-02 21:37:05 +00:00
|
|
|
* Copyright (c) 2016, 2018 The FreeBSD Foundation
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* Portions of this software were developed by Konstantin Belousov
|
|
|
|
* under sponsorship from the FreeBSD Foundation.
|
|
|
|
*
|
|
|
|
* Redistribution and use in source and binary forms, with or without
|
|
|
|
* modification, are permitted provided that the following conditions
|
|
|
|
* are met:
|
|
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer.
|
|
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
|
|
* documentation and/or other materials provided with the distribution.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
|
|
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
|
|
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
|
|
* SUCH DAMAGE.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <sys/cdefs.h>
|
|
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
|
|
|
|
#include <sys/param.h>
|
|
|
|
#include <sys/efi.h>
|
2018-12-15 05:46:04 +00:00
|
|
|
#include <sys/eventhandler.h>
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
#include <sys/kernel.h>
|
|
|
|
#include <sys/linker.h>
|
|
|
|
#include <sys/lock.h>
|
|
|
|
#include <sys/module.h>
|
|
|
|
#include <sys/mutex.h>
|
|
|
|
#include <sys/clock.h>
|
|
|
|
#include <sys/proc.h>
|
2018-12-15 05:46:04 +00:00
|
|
|
#include <sys/reboot.h>
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
#include <sys/rwlock.h>
|
|
|
|
#include <sys/sched.h>
|
|
|
|
#include <sys/sysctl.h>
|
|
|
|
#include <sys/systm.h>
|
2017-04-17 17:07:00 +00:00
|
|
|
#include <sys/vmmeter.h>
|
2017-10-01 19:52:47 +00:00
|
|
|
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
#include <machine/fpu.h>
|
|
|
|
#include <machine/efi.h>
|
|
|
|
#include <machine/metadata.h>
|
|
|
|
#include <machine/vmparam.h>
|
2017-10-01 19:52:47 +00:00
|
|
|
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
#include <vm/vm.h>
|
|
|
|
#include <vm/pmap.h>
|
2016-09-26 17:25:25 +00:00
|
|
|
#include <vm/vm_map.h>
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
|
|
|
|
static struct efi_systbl *efi_systbl;
|
2018-12-15 05:46:04 +00:00
|
|
|
static eventhandler_tag efi_shutdown_tag;
|
2018-03-23 02:45:09 +00:00
|
|
|
/*
|
|
|
|
* The following pointers point to tables in the EFI runtime service data pages.
|
|
|
|
* Care should be taken to make sure that we've properly entered the EFI runtime
|
|
|
|
* environment (efi_enter()) before dereferencing them.
|
|
|
|
*/
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
static struct efi_cfgtbl *efi_cfgtbl;
|
|
|
|
static struct efi_rt *efi_runtime;
|
|
|
|
|
|
|
|
static int efi_status2err[25] = {
|
|
|
|
0, /* EFI_SUCCESS */
|
|
|
|
ENOEXEC, /* EFI_LOAD_ERROR */
|
|
|
|
EINVAL, /* EFI_INVALID_PARAMETER */
|
|
|
|
ENOSYS, /* EFI_UNSUPPORTED */
|
|
|
|
EMSGSIZE, /* EFI_BAD_BUFFER_SIZE */
|
|
|
|
EOVERFLOW, /* EFI_BUFFER_TOO_SMALL */
|
|
|
|
EBUSY, /* EFI_NOT_READY */
|
|
|
|
EIO, /* EFI_DEVICE_ERROR */
|
|
|
|
EROFS, /* EFI_WRITE_PROTECTED */
|
|
|
|
EAGAIN, /* EFI_OUT_OF_RESOURCES */
|
|
|
|
EIO, /* EFI_VOLUME_CORRUPTED */
|
|
|
|
ENOSPC, /* EFI_VOLUME_FULL */
|
|
|
|
ENXIO, /* EFI_NO_MEDIA */
|
|
|
|
ESTALE, /* EFI_MEDIA_CHANGED */
|
|
|
|
ENOENT, /* EFI_NOT_FOUND */
|
|
|
|
EACCES, /* EFI_ACCESS_DENIED */
|
|
|
|
ETIMEDOUT, /* EFI_NO_RESPONSE */
|
|
|
|
EADDRNOTAVAIL, /* EFI_NO_MAPPING */
|
|
|
|
ETIMEDOUT, /* EFI_TIMEOUT */
|
|
|
|
EDOOFUS, /* EFI_NOT_STARTED */
|
|
|
|
EALREADY, /* EFI_ALREADY_STARTED */
|
|
|
|
ECANCELED, /* EFI_ABORTED */
|
|
|
|
EPROTO, /* EFI_ICMP_ERROR */
|
|
|
|
EPROTO, /* EFI_TFTP_ERROR */
|
|
|
|
EPROTO /* EFI_PROTOCOL_ERROR */
|
|
|
|
};
|
|
|
|
|
2018-03-22 15:32:57 +00:00
|
|
|
static int efi_enter(void);
|
|
|
|
static void efi_leave(void);
|
|
|
|
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
static int
|
|
|
|
efi_status_to_errno(efi_status status)
|
|
|
|
{
|
|
|
|
u_long code;
|
|
|
|
|
|
|
|
code = status & 0x3ffffffffffffffful;
|
|
|
|
return (code < nitems(efi_status2err) ? efi_status2err[code] : EDOOFUS);
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct mtx efi_lock;
|
2018-12-15 05:46:04 +00:00
|
|
|
static SYSCTL_NODE(_hw, OID_AUTO, efi, CTLFLAG_RWTUN, NULL, "EFI");
|
|
|
|
static bool efi_poweroff = true;
|
|
|
|
SYSCTL_BOOL(_hw_efi, OID_AUTO, poweroff, CTLFLAG_RWTUN, &efi_poweroff, 0,
|
|
|
|
"If true, use EFI runtime services to power off in preference to ACPI");
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
|
2018-03-20 13:35:20 +00:00
|
|
|
static bool
|
|
|
|
efi_is_in_map(struct efi_md *map, int ndesc, int descsz, vm_offset_t addr)
|
|
|
|
{
|
|
|
|
struct efi_md *p;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0, p = map; i < ndesc; i++, p = efi_next_descriptor(p,
|
|
|
|
descsz)) {
|
|
|
|
if ((p->md_attr & EFI_MD_ATTR_RT) == 0)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (addr >= (uintptr_t)p->md_virt &&
|
|
|
|
addr < (uintptr_t)p->md_virt + p->md_pages * PAGE_SIZE)
|
|
|
|
return (true);
|
|
|
|
}
|
|
|
|
|
|
|
|
return (false);
|
|
|
|
}
|
|
|
|
|
2018-12-15 05:46:04 +00:00
|
|
|
static void
|
|
|
|
efi_shutdown_final(void *dummy __unused, int howto)
|
|
|
|
{
|
|
|
|
|
|
|
|
/*
|
|
|
|
* On some systems, ACPI S5 is missing or does not function properly.
|
|
|
|
* When present, shutdown via EFI Runtime Services instead, unless
|
|
|
|
* disabled.
|
|
|
|
*/
|
|
|
|
if ((howto & RB_POWEROFF) != 0 && efi_poweroff)
|
|
|
|
(void)efi_reset_system(EFI_RESET_SHUTDOWN);
|
|
|
|
}
|
|
|
|
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
static int
|
|
|
|
efi_init(void)
|
|
|
|
{
|
|
|
|
struct efi_map_header *efihdr;
|
|
|
|
struct efi_md *map;
|
2018-08-04 21:41:10 +00:00
|
|
|
struct efi_rt *rtdm;
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
caddr_t kmdp;
|
|
|
|
size_t efisz;
|
2018-08-04 21:41:10 +00:00
|
|
|
int ndesc, rt_disabled;
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
|
2018-07-30 17:40:27 +00:00
|
|
|
rt_disabled = 0;
|
2018-07-30 18:13:20 +00:00
|
|
|
TUNABLE_INT_FETCH("efi.rt.disabled", &rt_disabled);
|
2018-07-30 17:40:27 +00:00
|
|
|
if (rt_disabled == 1)
|
|
|
|
return (0);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
mtx_init(&efi_lock, "efi", NULL, MTX_DEF);
|
|
|
|
|
|
|
|
if (efi_systbl_phys == 0) {
|
|
|
|
if (bootverbose)
|
|
|
|
printf("EFI systbl not available\n");
|
2016-12-09 23:37:11 +00:00
|
|
|
return (0);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
}
|
2018-08-04 21:41:10 +00:00
|
|
|
|
|
|
|
efi_systbl = (struct efi_systbl *)efi_phys_to_kva(efi_systbl_phys);
|
|
|
|
if (efi_systbl == NULL || efi_systbl->st_hdr.th_sig != EFI_SYSTBL_SIG) {
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
efi_systbl = NULL;
|
|
|
|
if (bootverbose)
|
|
|
|
printf("EFI systbl signature invalid\n");
|
2016-12-09 23:37:11 +00:00
|
|
|
return (0);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
}
|
|
|
|
efi_cfgtbl = (efi_systbl->st_cfgtbl == 0) ? NULL :
|
|
|
|
(struct efi_cfgtbl *)efi_systbl->st_cfgtbl;
|
|
|
|
if (efi_cfgtbl == NULL) {
|
|
|
|
if (bootverbose)
|
|
|
|
printf("EFI config table is not present\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
kmdp = preload_search_by_type("elf kernel");
|
|
|
|
if (kmdp == NULL)
|
|
|
|
kmdp = preload_search_by_type("elf64 kernel");
|
|
|
|
efihdr = (struct efi_map_header *)preload_search_info(kmdp,
|
|
|
|
MODINFO_METADATA | MODINFOMD_EFI_MAP);
|
|
|
|
if (efihdr == NULL) {
|
|
|
|
if (bootverbose)
|
|
|
|
printf("EFI map is not present\n");
|
2016-12-09 23:37:11 +00:00
|
|
|
return (0);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
}
|
|
|
|
efisz = (sizeof(struct efi_map_header) + 0xf) & ~0xf;
|
|
|
|
map = (struct efi_md *)((uint8_t *)efihdr + efisz);
|
|
|
|
if (efihdr->descriptor_size == 0)
|
|
|
|
return (ENOMEM);
|
|
|
|
|
2018-08-04 21:41:10 +00:00
|
|
|
ndesc = efihdr->memory_size / efihdr->descriptor_size;
|
|
|
|
if (!efi_create_1t1_map(map, ndesc, efihdr->descriptor_size)) {
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
if (bootverbose)
|
|
|
|
printf("EFI cannot create runtime map\n");
|
|
|
|
return (ENOMEM);
|
|
|
|
}
|
|
|
|
|
|
|
|
efi_runtime = (efi_systbl->st_rt == 0) ? NULL :
|
|
|
|
(struct efi_rt *)efi_systbl->st_rt;
|
|
|
|
if (efi_runtime == NULL) {
|
|
|
|
if (bootverbose)
|
|
|
|
printf("EFI runtime services table is not present\n");
|
|
|
|
efi_destroy_1t1_map();
|
|
|
|
return (ENXIO);
|
|
|
|
}
|
|
|
|
|
2018-08-04 21:41:10 +00:00
|
|
|
#if defined(__aarch64__) || defined(__amd64__)
|
2018-03-20 13:35:20 +00:00
|
|
|
/*
|
|
|
|
* Some UEFI implementations have multiple implementations of the
|
|
|
|
* RS->GetTime function. They switch from one we can only use early
|
|
|
|
* in the boot process to one valid as a RunTime service only when we
|
|
|
|
* call RS->SetVirtualAddressMap. As this is not always the case, e.g.
|
|
|
|
* with an old loader.efi, check if the RS->GetTime function is within
|
|
|
|
* the EFI map, and fail to attach if not.
|
|
|
|
*/
|
2018-08-04 21:41:10 +00:00
|
|
|
rtdm = (struct efi_rt *)efi_phys_to_kva((uintptr_t)efi_runtime);
|
|
|
|
if (rtdm == NULL || !efi_is_in_map(map, ndesc, efihdr->descriptor_size,
|
|
|
|
(vm_offset_t)rtdm->rt_gettime)) {
|
2018-03-20 13:35:20 +00:00
|
|
|
if (bootverbose)
|
|
|
|
printf(
|
|
|
|
"EFI runtime services table has an invalid pointer\n");
|
|
|
|
efi_runtime = NULL;
|
|
|
|
efi_destroy_1t1_map();
|
|
|
|
return (ENXIO);
|
|
|
|
}
|
2018-08-04 21:41:10 +00:00
|
|
|
#endif
|
2018-03-20 13:35:20 +00:00
|
|
|
|
2018-12-15 05:46:04 +00:00
|
|
|
/*
|
|
|
|
* We use SHUTDOWN_PRI_LAST - 1 to trigger after IPMI, but before ACPI.
|
|
|
|
*/
|
|
|
|
efi_shutdown_tag = EVENTHANDLER_REGISTER(shutdown_final,
|
|
|
|
efi_shutdown_final, NULL, SHUTDOWN_PRI_LAST - 1);
|
|
|
|
|
2016-10-16 06:07:43 +00:00
|
|
|
return (0);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
efi_uninit(void)
|
|
|
|
{
|
|
|
|
|
2018-07-30 17:40:27 +00:00
|
|
|
/* Most likely disabled by tunable */
|
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return;
|
2018-12-15 05:46:04 +00:00
|
|
|
if (efi_shutdown_tag != NULL)
|
|
|
|
EVENTHANDLER_DEREGISTER(shutdown_final, efi_shutdown_tag);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
efi_destroy_1t1_map();
|
|
|
|
|
|
|
|
efi_systbl = NULL;
|
|
|
|
efi_cfgtbl = NULL;
|
|
|
|
efi_runtime = NULL;
|
|
|
|
|
|
|
|
mtx_destroy(&efi_lock);
|
|
|
|
}
|
|
|
|
|
2017-08-08 20:44:16 +00:00
|
|
|
int
|
|
|
|
efi_rt_ok(void)
|
|
|
|
{
|
|
|
|
|
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
2017-10-01 19:52:47 +00:00
|
|
|
static int
|
|
|
|
efi_enter(void)
|
|
|
|
{
|
|
|
|
struct thread *td;
|
|
|
|
pmap_t curpmap;
|
2019-06-03 15:41:45 +00:00
|
|
|
int error;
|
2017-10-01 19:52:47 +00:00
|
|
|
|
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
td = curthread;
|
|
|
|
curpmap = &td->td_proc->p_vmspace->vm_pmap;
|
|
|
|
PMAP_LOCK(curpmap);
|
|
|
|
mtx_lock(&efi_lock);
|
2018-02-23 20:15:19 +00:00
|
|
|
fpu_kern_enter(td, NULL, FPU_KERN_NOCTX);
|
2019-06-03 15:41:45 +00:00
|
|
|
error = efi_arch_enter();
|
|
|
|
if (error != 0) {
|
|
|
|
fpu_kern_leave(td, NULL);
|
|
|
|
mtx_unlock(&efi_lock);
|
|
|
|
PMAP_UNLOCK(curpmap);
|
|
|
|
}
|
|
|
|
return (error);
|
2017-10-01 19:52:47 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
efi_leave(void)
|
|
|
|
{
|
|
|
|
struct thread *td;
|
|
|
|
pmap_t curpmap;
|
|
|
|
|
|
|
|
efi_arch_leave();
|
|
|
|
|
|
|
|
curpmap = &curproc->p_vmspace->vm_pmap;
|
|
|
|
td = curthread;
|
|
|
|
fpu_kern_leave(td, NULL);
|
|
|
|
mtx_unlock(&efi_lock);
|
|
|
|
PMAP_UNLOCK(curpmap);
|
|
|
|
}
|
|
|
|
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
int
|
2016-09-22 19:04:51 +00:00
|
|
|
efi_get_table(struct uuid *uuid, void **ptr)
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
{
|
|
|
|
struct efi_cfgtbl *ct;
|
|
|
|
u_long count;
|
|
|
|
|
2017-08-08 20:44:16 +00:00
|
|
|
if (efi_cfgtbl == NULL || efi_systbl == NULL)
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
return (ENXIO);
|
|
|
|
count = efi_systbl->st_entries;
|
|
|
|
ct = efi_cfgtbl;
|
|
|
|
while (count--) {
|
|
|
|
if (!bcmp(&ct->ct_uuid, uuid, sizeof(*uuid))) {
|
2018-08-04 21:41:10 +00:00
|
|
|
*ptr = (void *)efi_phys_to_kva(ct->ct_data);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
ct++;
|
|
|
|
}
|
|
|
|
return (ENOENT);
|
|
|
|
}
|
|
|
|
|
2018-09-02 21:37:05 +00:00
|
|
|
static int efi_rt_handle_faults = EFI_RT_HANDLE_FAULTS_DEFAULT;
|
|
|
|
SYSCTL_INT(_machdep, OID_AUTO, efi_rt_handle_faults, CTLFLAG_RWTUN,
|
|
|
|
&efi_rt_handle_faults, 0,
|
|
|
|
"Call EFI RT methods with fault handler wrapper around");
|
|
|
|
|
2017-10-01 19:52:47 +00:00
|
|
|
static int
|
2018-09-02 21:37:05 +00:00
|
|
|
efi_rt_arch_call_nofault(struct efirt_callinfo *ec)
|
|
|
|
{
|
|
|
|
|
|
|
|
switch (ec->ec_argcnt) {
|
|
|
|
case 0:
|
|
|
|
ec->ec_efi_status = ((register_t (*)(void))ec->ec_fptr)();
|
|
|
|
break;
|
|
|
|
case 1:
|
|
|
|
ec->ec_efi_status = ((register_t (*)(register_t))ec->ec_fptr)
|
|
|
|
(ec->ec_arg1);
|
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
ec->ec_efi_status = ((register_t (*)(register_t, register_t))
|
|
|
|
ec->ec_fptr)(ec->ec_arg1, ec->ec_arg2);
|
|
|
|
break;
|
|
|
|
case 3:
|
|
|
|
ec->ec_efi_status = ((register_t (*)(register_t, register_t,
|
|
|
|
register_t))ec->ec_fptr)(ec->ec_arg1, ec->ec_arg2,
|
|
|
|
ec->ec_arg3);
|
|
|
|
break;
|
|
|
|
case 4:
|
|
|
|
ec->ec_efi_status = ((register_t (*)(register_t, register_t,
|
|
|
|
register_t, register_t))ec->ec_fptr)(ec->ec_arg1,
|
|
|
|
ec->ec_arg2, ec->ec_arg3, ec->ec_arg4);
|
|
|
|
break;
|
|
|
|
case 5:
|
|
|
|
ec->ec_efi_status = ((register_t (*)(register_t, register_t,
|
|
|
|
register_t, register_t, register_t))ec->ec_fptr)(
|
|
|
|
ec->ec_arg1, ec->ec_arg2, ec->ec_arg3, ec->ec_arg4,
|
|
|
|
ec->ec_arg5);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
panic("efi_rt_arch_call: %d args", (int)ec->ec_argcnt);
|
|
|
|
}
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
efi_call(struct efirt_callinfo *ecp)
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
error = efi_enter();
|
|
|
|
if (error != 0)
|
|
|
|
return (error);
|
2018-09-02 21:37:05 +00:00
|
|
|
error = efi_rt_handle_faults ? efi_rt_arch_call(ecp) :
|
|
|
|
efi_rt_arch_call_nofault(ecp);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
efi_leave();
|
2018-09-02 21:37:05 +00:00
|
|
|
if (error == 0)
|
|
|
|
error = efi_status_to_errno(ecp->ec_efi_status);
|
|
|
|
else if (bootverbose)
|
|
|
|
printf("EFI %s call faulted, error %d\n", ecp->ec_name, error);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
2018-09-02 21:37:05 +00:00
|
|
|
#define EFI_RT_METHOD_PA(method) \
|
|
|
|
((uintptr_t)((struct efi_rt *)efi_phys_to_kva((uintptr_t) \
|
|
|
|
efi_runtime))->method)
|
|
|
|
|
|
|
|
static int
|
|
|
|
efi_get_time_locked(struct efi_tm *tm, struct efi_tmcap *tmcap)
|
|
|
|
{
|
|
|
|
struct efirt_callinfo ec;
|
|
|
|
|
|
|
|
EFI_TIME_OWNED();
|
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
bzero(&ec, sizeof(ec));
|
|
|
|
ec.ec_name = "rt_gettime";
|
|
|
|
ec.ec_argcnt = 2;
|
|
|
|
ec.ec_arg1 = (uintptr_t)tm;
|
|
|
|
ec.ec_arg2 = (uintptr_t)tmcap;
|
|
|
|
ec.ec_fptr = EFI_RT_METHOD_PA(rt_gettime);
|
|
|
|
return (efi_call(&ec));
|
|
|
|
}
|
|
|
|
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
int
|
|
|
|
efi_get_time(struct efi_tm *tm)
|
|
|
|
{
|
2018-03-13 15:01:23 +00:00
|
|
|
struct efi_tmcap dummy;
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
int error;
|
|
|
|
|
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
2018-09-02 19:48:41 +00:00
|
|
|
EFI_TIME_LOCK();
|
2018-03-13 15:01:23 +00:00
|
|
|
/*
|
|
|
|
* UEFI spec states that the Capabilities argument to GetTime is
|
|
|
|
* optional, but some UEFI implementations choke when passed a NULL
|
2018-03-13 15:02:46 +00:00
|
|
|
* pointer. Pass a dummy efi_tmcap, even though we won't use it,
|
2018-03-13 15:01:23 +00:00
|
|
|
* to workaround such implementations.
|
|
|
|
*/
|
|
|
|
error = efi_get_time_locked(tm, &dummy);
|
2018-09-02 19:48:41 +00:00
|
|
|
EFI_TIME_UNLOCK();
|
2018-03-01 22:57:14 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
efi_get_time_capabilities(struct efi_tmcap *tmcap)
|
|
|
|
{
|
|
|
|
struct efi_tm dummy;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
2018-09-02 19:48:41 +00:00
|
|
|
EFI_TIME_LOCK();
|
2018-03-01 22:57:14 +00:00
|
|
|
error = efi_get_time_locked(&dummy, tmcap);
|
2018-09-02 19:48:41 +00:00
|
|
|
EFI_TIME_UNLOCK();
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
2018-12-15 05:46:04 +00:00
|
|
|
efi_reset_system(enum efi_reset type)
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
{
|
2018-09-02 21:37:05 +00:00
|
|
|
struct efirt_callinfo ec;
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
|
2018-12-15 05:46:04 +00:00
|
|
|
switch (type) {
|
|
|
|
case EFI_RESET_COLD:
|
|
|
|
case EFI_RESET_WARM:
|
|
|
|
case EFI_RESET_SHUTDOWN:
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
2018-09-02 21:37:05 +00:00
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
bzero(&ec, sizeof(ec));
|
|
|
|
ec.ec_name = "rt_reset";
|
|
|
|
ec.ec_argcnt = 4;
|
2018-12-15 05:46:04 +00:00
|
|
|
ec.ec_arg1 = (uintptr_t)type;
|
2018-09-04 19:28:46 +00:00
|
|
|
ec.ec_arg2 = (uintptr_t)0;
|
|
|
|
ec.ec_arg3 = (uintptr_t)0;
|
|
|
|
ec.ec_arg4 = (uintptr_t)NULL;
|
2018-09-02 21:37:05 +00:00
|
|
|
ec.ec_fptr = EFI_RT_METHOD_PA(rt_reset);
|
|
|
|
return (efi_call(&ec));
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
}
|
|
|
|
|
2017-10-01 19:52:47 +00:00
|
|
|
static int
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
efi_set_time_locked(struct efi_tm *tm)
|
|
|
|
{
|
2018-09-02 21:37:05 +00:00
|
|
|
struct efirt_callinfo ec;
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
|
2017-10-01 19:52:47 +00:00
|
|
|
EFI_TIME_OWNED();
|
2018-09-02 21:37:05 +00:00
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
bzero(&ec, sizeof(ec));
|
|
|
|
ec.ec_name = "rt_settime";
|
|
|
|
ec.ec_argcnt = 1;
|
|
|
|
ec.ec_arg1 = (uintptr_t)tm;
|
|
|
|
ec.ec_fptr = EFI_RT_METHOD_PA(rt_settime);
|
|
|
|
return (efi_call(&ec));
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
efi_set_time(struct efi_tm *tm)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
2018-09-02 19:48:41 +00:00
|
|
|
EFI_TIME_LOCK();
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
error = efi_set_time_locked(tm);
|
2018-09-02 19:48:41 +00:00
|
|
|
EFI_TIME_UNLOCK();
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
efi_var_get(efi_char *name, struct uuid *vendor, uint32_t *attrib,
|
|
|
|
size_t *datasize, void *data)
|
|
|
|
{
|
2018-09-02 21:37:05 +00:00
|
|
|
struct efirt_callinfo ec;
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
|
2018-09-02 21:37:05 +00:00
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
bzero(&ec, sizeof(ec));
|
|
|
|
ec.ec_argcnt = 5;
|
|
|
|
ec.ec_name = "rt_getvar";
|
|
|
|
ec.ec_arg1 = (uintptr_t)name;
|
|
|
|
ec.ec_arg2 = (uintptr_t)vendor;
|
|
|
|
ec.ec_arg3 = (uintptr_t)attrib;
|
|
|
|
ec.ec_arg4 = (uintptr_t)datasize;
|
|
|
|
ec.ec_arg5 = (uintptr_t)data;
|
|
|
|
ec.ec_fptr = EFI_RT_METHOD_PA(rt_getvar);
|
|
|
|
return (efi_call(&ec));
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
efi_var_nextname(size_t *namesize, efi_char *name, struct uuid *vendor)
|
|
|
|
{
|
2018-09-02 21:37:05 +00:00
|
|
|
struct efirt_callinfo ec;
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
|
2018-09-02 21:37:05 +00:00
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
bzero(&ec, sizeof(ec));
|
|
|
|
ec.ec_argcnt = 3;
|
|
|
|
ec.ec_name = "rt_scanvar";
|
|
|
|
ec.ec_arg1 = (uintptr_t)namesize;
|
|
|
|
ec.ec_arg2 = (uintptr_t)name;
|
|
|
|
ec.ec_arg3 = (uintptr_t)vendor;
|
|
|
|
ec.ec_fptr = EFI_RT_METHOD_PA(rt_scanvar);
|
|
|
|
return (efi_call(&ec));
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
efi_var_set(efi_char *name, struct uuid *vendor, uint32_t attrib,
|
|
|
|
size_t datasize, void *data)
|
|
|
|
{
|
2018-09-02 21:37:05 +00:00
|
|
|
struct efirt_callinfo ec;
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
|
2018-09-02 21:37:05 +00:00
|
|
|
if (efi_runtime == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
bzero(&ec, sizeof(ec));
|
|
|
|
ec.ec_argcnt = 5;
|
|
|
|
ec.ec_name = "rt_setvar";
|
|
|
|
ec.ec_arg1 = (uintptr_t)name;
|
|
|
|
ec.ec_arg2 = (uintptr_t)vendor;
|
|
|
|
ec.ec_arg3 = (uintptr_t)attrib;
|
|
|
|
ec.ec_arg4 = (uintptr_t)datasize;
|
|
|
|
ec.ec_arg5 = (uintptr_t)data;
|
|
|
|
ec.ec_fptr = EFI_RT_METHOD_PA(rt_setvar);
|
|
|
|
return (efi_call(&ec));
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
efirt_modevents(module_t m, int event, void *arg __unused)
|
|
|
|
{
|
|
|
|
|
|
|
|
switch (event) {
|
|
|
|
case MOD_LOAD:
|
|
|
|
return (efi_init());
|
|
|
|
|
|
|
|
case MOD_UNLOAD:
|
|
|
|
efi_uninit();
|
|
|
|
return (0);
|
|
|
|
|
|
|
|
case MOD_SHUTDOWN:
|
|
|
|
return (0);
|
|
|
|
|
|
|
|
default:
|
|
|
|
return (EOPNOTSUPP);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static moduledata_t efirt_moddata = {
|
|
|
|
.name = "efirt",
|
|
|
|
.evhand = efirt_modevents,
|
|
|
|
.priv = NULL,
|
|
|
|
};
|
2018-03-22 18:24:00 +00:00
|
|
|
/* After fpuinitstate, before efidev */
|
|
|
|
DECLARE_MODULE(efirt, efirt_moddata, SI_SUB_DRIVERS, SI_ORDER_SECOND);
|
Add kernel interfaces to call EFI Runtime Services.
Runtime services require special execution environment for the call.
Besides that, OS must inform firmware about runtime virtual memory map
which will be active during the calls, with the SetVirtualAddressMap()
runtime call, done while the 1:1 mapping is still used. There are two
complication: the SetVirtualAddressMap() effectively must be done from
loader, which needs to know kernel address map in advance. More,
despite not explicitely mentioned in the specification, both 1:1 and
the map passed to SetVirtualAddressMap() must be active during the
SetVirtualAddressMap() call. Second, there are buggy BIOSes which
require both mappings active during runtime calls as well, most likely
because they fail to identify all relocations to perform.
On amd64, we can get rid of both problems by providing 1:1 mapping for
the duration of runtime calls, by temprorary remapping user addresses.
As result, we avoid the need for loader to know about future kernel
address map, and avoid bugs in BIOSes. Typically BIOS only maps
something in low 4G. If not runtime bugs, we would take advantage of
the DMAP, as previous versions of this patch did.
Similar but more complicated trick can be used even for i386 and 32bit
runtime, if and when the EFI boot on i386 is supported. We would need
a trampoline page, since potentially whole 4G of VA would be switched
on calls, instead of only userspace portion on amd64.
Context switches are disabled for the duration of the call, FPU access
is granted, and interrupts are not disabled. The later is possible
because kernel is mapped during calls.
To test, the sysctl mib debug.efi_time is provided, setting it to 1
makes one call to EFI get_time() runtime service, on success the efitm
structure is printed to the control terminal. Load efirt.ko, or add
EFIRT option to the kernel config, to enable code.
Discussed with: emaste, imp
Tested by: emaste (mac, qemu)
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
2016-09-21 11:31:58 +00:00
|
|
|
MODULE_VERSION(efirt, 1);
|