2008-07-29 17:29:19 +00:00
|
|
|
/*
|
|
|
|
* xsave/xrstor support.
|
|
|
|
*
|
|
|
|
* Author: Suresh Siddha <suresh.b.siddha@intel.com>
|
|
|
|
*/
|
|
|
|
#include <linux/bootmem.h>
|
|
|
|
#include <linux/compat.h>
|
|
|
|
#include <asm/i387.h>
|
2008-07-29 17:29:25 +00:00
|
|
|
#ifdef CONFIG_IA32_EMULATION
|
|
|
|
#include <asm/sigcontext32.h>
|
|
|
|
#endif
|
2008-07-30 00:23:16 +00:00
|
|
|
#include <asm/xcr.h>
|
2008-07-29 17:29:19 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Supported feature mask by the CPU and the kernel.
|
|
|
|
*/
|
2008-07-30 00:23:16 +00:00
|
|
|
u64 pcntxt_mask;
|
2008-07-29 17:29:19 +00:00
|
|
|
|
2010-07-21 17:03:55 +00:00
|
|
|
/*
|
|
|
|
* Represents init state for the supported extended state.
|
|
|
|
*/
|
|
|
|
static struct xsave_struct *init_xstate_buf;
|
|
|
|
|
2008-07-29 17:29:25 +00:00
|
|
|
struct _fpx_sw_bytes fx_sw_reserved;
|
|
|
|
#ifdef CONFIG_IA32_EMULATION
|
|
|
|
struct _fpx_sw_bytes fx_sw_reserved_ia32;
|
|
|
|
#endif
|
|
|
|
|
2010-07-19 23:05:48 +00:00
|
|
|
static unsigned int *xstate_offsets, *xstate_sizes, xstate_features;
|
|
|
|
|
2010-07-19 23:05:49 +00:00
|
|
|
/*
|
|
|
|
* If a processor implementation discern that a processor state component is
|
|
|
|
* in its initialized state it may modify the corresponding bit in the
|
|
|
|
* xsave_hdr.xstate_bv as '0', with out modifying the corresponding memory
|
|
|
|
* layout in the case of xsaveopt. While presenting the xstate information to
|
|
|
|
* the user, we always ensure that the memory layout of a feature will be in
|
|
|
|
* the init state if the corresponding header bit is zero. This is to ensure
|
|
|
|
* that the user doesn't see some stale state in the memory layout during
|
|
|
|
* signal handling, debugging etc.
|
|
|
|
*/
|
|
|
|
void __sanitize_i387_state(struct task_struct *tsk)
|
|
|
|
{
|
|
|
|
u64 xstate_bv;
|
|
|
|
int feature_bit = 0x2;
|
|
|
|
struct i387_fxsave_struct *fx = &tsk->thread.fpu.state->fxsave;
|
|
|
|
|
|
|
|
if (!fx)
|
|
|
|
return;
|
|
|
|
|
i387: move TS_USEDFPU flag from thread_info to task_struct
This moves the bit that indicates whether a thread has ownership of the
FPU from the TS_USEDFPU bit in thread_info->status to a word of its own
(called 'has_fpu') in task_struct->thread.has_fpu.
This fixes two independent bugs at the same time:
- changing 'thread_info->status' from the scheduler causes nasty
problems for the other users of that variable, since it is defined to
be thread-synchronous (that's what the "TS_" part of the naming was
supposed to indicate).
So perfectly valid code could (and did) do
ti->status |= TS_RESTORE_SIGMASK;
and the compiler was free to do that as separate load, or and store
instructions. Which can cause problems with preemption, since a task
switch could happen in between, and change the TS_USEDFPU bit. The
change to TS_USEDFPU would be overwritten by the final store.
In practice, this seldom happened, though, because the 'status' field
was seldom used more than once, so gcc would generally tend to
generate code that used a read-modify-write instruction and thus
happened to avoid this problem - RMW instructions are naturally low
fat and preemption-safe.
- On x86-32, the current_thread_info() pointer would, during interrupts
and softirqs, point to a *copy* of the real thread_info, because
x86-32 uses %esp to calculate the thread_info address, and thus the
separate irq (and softirq) stacks would cause these kinds of odd
thread_info copy aliases.
This is normally not a problem, since interrupts aren't supposed to
look at thread information anyway (what thread is running at
interrupt time really isn't very well-defined), but it confused the
heck out of irq_fpu_usable() and the code that tried to squirrel
away the FPU state.
(It also caused untold confusion for us poor kernel developers).
It also turns out that using 'task_struct' is actually much more natural
for most of the call sites that care about the FPU state, since they
tend to work with the task struct for other reasons anyway (ie
scheduling). And the FPU data that we are going to save/restore is
found there too.
Thanks to Arjan Van De Ven <arjan@linux.intel.com> for pointing us to
the %esp issue.
Cc: Arjan van de Ven <arjan@linux.intel.com>
Reported-and-tested-by: Raphael Prevost <raphael@buro.asia>
Acked-and-tested-by: Suresh Siddha <suresh.b.siddha@intel.com>
Tested-by: Peter Anvin <hpa@zytor.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-02-18 05:48:54 +00:00
|
|
|
BUG_ON(__thread_has_fpu(tsk));
|
2010-07-19 23:05:49 +00:00
|
|
|
|
|
|
|
xstate_bv = tsk->thread.fpu.state->xsave.xsave_hdr.xstate_bv;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* None of the feature bits are in init state. So nothing else
|
2011-03-17 19:24:16 +00:00
|
|
|
* to do for us, as the memory layout is up to date.
|
2010-07-19 23:05:49 +00:00
|
|
|
*/
|
|
|
|
if ((xstate_bv & pcntxt_mask) == pcntxt_mask)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* FP is in init state
|
|
|
|
*/
|
|
|
|
if (!(xstate_bv & XSTATE_FP)) {
|
|
|
|
fx->cwd = 0x37f;
|
|
|
|
fx->swd = 0;
|
|
|
|
fx->twd = 0;
|
|
|
|
fx->fop = 0;
|
|
|
|
fx->rip = 0;
|
|
|
|
fx->rdp = 0;
|
|
|
|
memset(&fx->st_space[0], 0, 128);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* SSE is in init state
|
|
|
|
*/
|
|
|
|
if (!(xstate_bv & XSTATE_SSE))
|
|
|
|
memset(&fx->xmm_space[0], 0, 256);
|
|
|
|
|
|
|
|
xstate_bv = (pcntxt_mask & ~xstate_bv) >> 2;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update all the other memory layouts for which the corresponding
|
|
|
|
* header bit is in the init state.
|
|
|
|
*/
|
|
|
|
while (xstate_bv) {
|
|
|
|
if (xstate_bv & 0x1) {
|
|
|
|
int offset = xstate_offsets[feature_bit];
|
|
|
|
int size = xstate_sizes[feature_bit];
|
|
|
|
|
|
|
|
memcpy(((void *) fx) + offset,
|
|
|
|
((void *) init_xstate_buf) + offset,
|
|
|
|
size);
|
|
|
|
}
|
|
|
|
|
|
|
|
xstate_bv >>= 1;
|
|
|
|
feature_bit++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-07-29 17:29:25 +00:00
|
|
|
/*
|
|
|
|
* Check for the presence of extended state information in the
|
|
|
|
* user fpstate pointer in the sigcontext.
|
|
|
|
*/
|
|
|
|
int check_for_xstate(struct i387_fxsave_struct __user *buf,
|
|
|
|
void __user *fpstate,
|
|
|
|
struct _fpx_sw_bytes *fx_sw_user)
|
|
|
|
{
|
|
|
|
int min_xstate_size = sizeof(struct i387_fxsave_struct) +
|
|
|
|
sizeof(struct xsave_hdr_struct);
|
|
|
|
unsigned int magic2;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = __copy_from_user(fx_sw_user, &buf->sw_reserved[0],
|
|
|
|
sizeof(struct _fpx_sw_bytes));
|
|
|
|
if (err)
|
2010-06-03 10:07:46 +00:00
|
|
|
return -EFAULT;
|
2008-07-29 17:29:25 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* First Magic check failed.
|
|
|
|
*/
|
|
|
|
if (fx_sw_user->magic1 != FP_XSTATE_MAGIC1)
|
2010-06-03 10:07:46 +00:00
|
|
|
return -EINVAL;
|
2008-07-29 17:29:25 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Check for error scenarios.
|
|
|
|
*/
|
|
|
|
if (fx_sw_user->xstate_size < min_xstate_size ||
|
|
|
|
fx_sw_user->xstate_size > xstate_size ||
|
|
|
|
fx_sw_user->xstate_size > fx_sw_user->extended_size)
|
2010-06-03 10:07:46 +00:00
|
|
|
return -EINVAL;
|
2008-07-29 17:29:25 +00:00
|
|
|
|
|
|
|
err = __get_user(magic2, (__u32 *) (((void *)fpstate) +
|
|
|
|
fx_sw_user->extended_size -
|
|
|
|
FP_XSTATE_MAGIC2_SIZE));
|
2010-06-03 10:07:46 +00:00
|
|
|
if (err)
|
|
|
|
return err;
|
2008-07-29 17:29:25 +00:00
|
|
|
/*
|
|
|
|
* Check for the presence of second magic word at the end of memory
|
|
|
|
* layout. This detects the case where the user just copied the legacy
|
|
|
|
* fpstate layout with out copying the extended state information
|
|
|
|
* in the memory layout.
|
|
|
|
*/
|
2010-06-03 10:07:46 +00:00
|
|
|
if (magic2 != FP_XSTATE_MAGIC2)
|
|
|
|
return -EFAULT;
|
2008-07-29 17:29:25 +00:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-07-29 17:29:22 +00:00
|
|
|
#ifdef CONFIG_X86_64
|
|
|
|
/*
|
|
|
|
* Signal frame handlers.
|
|
|
|
*/
|
|
|
|
|
|
|
|
int save_i387_xstate(void __user *buf)
|
|
|
|
{
|
|
|
|
struct task_struct *tsk = current;
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
if (!access_ok(VERIFY_WRITE, buf, sig_xstate_size))
|
|
|
|
return -EACCES;
|
|
|
|
|
2008-08-13 18:38:15 +00:00
|
|
|
BUG_ON(sig_xstate_size < xstate_size);
|
2008-07-29 17:29:22 +00:00
|
|
|
|
2008-07-29 17:29:25 +00:00
|
|
|
if ((unsigned long)buf % 64)
|
2008-07-29 17:29:22 +00:00
|
|
|
printk("save_i387_xstate: bad fpstate %p\n", buf);
|
|
|
|
|
|
|
|
if (!used_math())
|
|
|
|
return 0;
|
2009-04-09 22:24:34 +00:00
|
|
|
|
i387: fix x86-64 preemption-unsafe user stack save/restore
Commit 5b1cbac37798 ("i387: make irq_fpu_usable() tests more robust")
added a sanity check to the #NM handler to verify that we never cause
the "Device Not Available" exception in kernel mode.
However, that check actually pinpointed a (fundamental) race where we do
cause that exception as part of the signal stack FPU state save/restore
code.
Because we use the floating point instructions themselves to save and
restore state directly from user mode, we cannot do that atomically with
testing the TS_USEDFPU bit: the user mode access itself may cause a page
fault, which causes a task switch, which saves and restores the FP/MMX
state from the kernel buffers.
This kind of "recursive" FP state save is fine per se, but it means that
when the signal stack save/restore gets restarted, it will now take the
'#NM' exception we originally tried to avoid. With preemption this can
happen even without the page fault - but because of the user access, we
cannot just disable preemption around the save/restore instruction.
There are various ways to solve this, including using the
"enable/disable_page_fault()" helpers to not allow page faults at all
during the sequence, and fall back to copying things by hand without the
use of the native FP state save/restore instructions.
However, the simplest thing to do is to just allow the #NM from kernel
space, but fix the race in setting and clearing CR0.TS that this all
exposed: the TS bit changes and the TS_USEDFPU bit absolutely have to be
atomic wrt scheduling, so while the actual state save/restore can be
interrupted and restarted, the act of actually clearing/setting CR0.TS
and the TS_USEDFPU bit together must not.
Instead of just adding random "preempt_disable/enable()" calls to what
is already excessively ugly code, this introduces some helper functions
that mostly mirror the "kernel_fpu_begin/end()" functionality, just for
the user state instead.
Those helper functions should probably eventually replace the other
ad-hoc CR0.TS and TS_USEDFPU tests too, but I'll need to think about it
some more: the task switching functionality in particular needs to
expose the difference between the 'prev' and 'next' threads, while the
new helper functions intentionally were written to only work with
'current'.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-02-16 17:15:04 +00:00
|
|
|
if (user_has_fpu()) {
|
2010-05-06 08:45:45 +00:00
|
|
|
if (use_xsave())
|
2008-07-29 17:29:25 +00:00
|
|
|
err = xsave_user(buf);
|
|
|
|
else
|
|
|
|
err = fxsave_user(buf);
|
|
|
|
|
2008-07-29 17:29:22 +00:00
|
|
|
if (err)
|
|
|
|
return err;
|
i387: fix x86-64 preemption-unsafe user stack save/restore
Commit 5b1cbac37798 ("i387: make irq_fpu_usable() tests more robust")
added a sanity check to the #NM handler to verify that we never cause
the "Device Not Available" exception in kernel mode.
However, that check actually pinpointed a (fundamental) race where we do
cause that exception as part of the signal stack FPU state save/restore
code.
Because we use the floating point instructions themselves to save and
restore state directly from user mode, we cannot do that atomically with
testing the TS_USEDFPU bit: the user mode access itself may cause a page
fault, which causes a task switch, which saves and restores the FP/MMX
state from the kernel buffers.
This kind of "recursive" FP state save is fine per se, but it means that
when the signal stack save/restore gets restarted, it will now take the
'#NM' exception we originally tried to avoid. With preemption this can
happen even without the page fault - but because of the user access, we
cannot just disable preemption around the save/restore instruction.
There are various ways to solve this, including using the
"enable/disable_page_fault()" helpers to not allow page faults at all
during the sequence, and fall back to copying things by hand without the
use of the native FP state save/restore instructions.
However, the simplest thing to do is to just allow the #NM from kernel
space, but fix the race in setting and clearing CR0.TS that this all
exposed: the TS bit changes and the TS_USEDFPU bit absolutely have to be
atomic wrt scheduling, so while the actual state save/restore can be
interrupted and restarted, the act of actually clearing/setting CR0.TS
and the TS_USEDFPU bit together must not.
Instead of just adding random "preempt_disable/enable()" calls to what
is already excessively ugly code, this introduces some helper functions
that mostly mirror the "kernel_fpu_begin/end()" functionality, just for
the user state instead.
Those helper functions should probably eventually replace the other
ad-hoc CR0.TS and TS_USEDFPU tests too, but I'll need to think about it
some more: the task switching functionality in particular needs to
expose the difference between the 'prev' and 'next' threads, while the
new helper functions intentionally were written to only work with
'current'.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-02-16 17:15:04 +00:00
|
|
|
user_fpu_end();
|
2008-07-29 17:29:22 +00:00
|
|
|
} else {
|
2010-07-19 23:05:49 +00:00
|
|
|
sanitize_i387_state(tsk);
|
2010-05-06 08:45:46 +00:00
|
|
|
if (__copy_to_user(buf, &tsk->thread.fpu.state->fxsave,
|
2008-07-29 17:29:22 +00:00
|
|
|
xstate_size))
|
|
|
|
return -1;
|
|
|
|
}
|
2008-07-29 17:29:25 +00:00
|
|
|
|
2009-04-09 22:24:34 +00:00
|
|
|
clear_used_math(); /* trigger finit */
|
|
|
|
|
2010-05-06 08:45:45 +00:00
|
|
|
if (use_xsave()) {
|
2008-07-29 17:29:25 +00:00
|
|
|
struct _fpstate __user *fx = buf;
|
2008-10-07 21:04:28 +00:00
|
|
|
struct _xstate __user *x = buf;
|
|
|
|
u64 xstate_bv;
|
2008-07-29 17:29:25 +00:00
|
|
|
|
|
|
|
err = __copy_to_user(&fx->sw_reserved, &fx_sw_reserved,
|
|
|
|
sizeof(struct _fpx_sw_bytes));
|
|
|
|
|
|
|
|
err |= __put_user(FP_XSTATE_MAGIC2,
|
|
|
|
(__u32 __user *) (buf + sig_xstate_size
|
|
|
|
- FP_XSTATE_MAGIC2_SIZE));
|
2008-10-07 21:04:28 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Read the xstate_bv which we copied (directly from the cpu or
|
|
|
|
* from the state in task struct) to the user buffers and
|
|
|
|
* set the FP/SSE bits.
|
|
|
|
*/
|
|
|
|
err |= __get_user(xstate_bv, &x->xstate_hdr.xstate_bv);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For legacy compatible, we always set FP/SSE bits in the bit
|
|
|
|
* vector while saving the state to the user context. This will
|
|
|
|
* enable us capturing any changes(during sigreturn) to
|
|
|
|
* the FP/SSE bits by the legacy applications which don't touch
|
|
|
|
* xstate_bv in the xsave header.
|
|
|
|
*
|
|
|
|
* xsave aware apps can change the xstate_bv in the xsave
|
|
|
|
* header as well as change any contents in the memory layout.
|
|
|
|
* xrestore as part of sigreturn will capture all the changes.
|
|
|
|
*/
|
|
|
|
xstate_bv |= XSTATE_FPSSE;
|
|
|
|
|
|
|
|
err |= __put_user(xstate_bv, &x->xstate_hdr.xstate_bv);
|
|
|
|
|
2008-10-07 21:04:27 +00:00
|
|
|
if (err)
|
|
|
|
return err;
|
2008-07-29 17:29:25 +00:00
|
|
|
}
|
|
|
|
|
2008-07-29 17:29:22 +00:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2008-07-29 17:29:25 +00:00
|
|
|
/*
|
|
|
|
* Restore the extended state if present. Otherwise, restore the FP/SSE
|
|
|
|
* state.
|
|
|
|
*/
|
2008-12-30 16:35:55 +00:00
|
|
|
static int restore_user_xstate(void __user *buf)
|
2008-07-29 17:29:25 +00:00
|
|
|
{
|
|
|
|
struct _fpx_sw_bytes fx_sw_user;
|
2008-07-30 00:23:16 +00:00
|
|
|
u64 mask;
|
2008-07-29 17:29:25 +00:00
|
|
|
int err;
|
|
|
|
|
|
|
|
if (((unsigned long)buf % 64) ||
|
|
|
|
check_for_xstate(buf, buf, &fx_sw_user))
|
|
|
|
goto fx_only;
|
|
|
|
|
2008-07-30 00:23:16 +00:00
|
|
|
mask = fx_sw_user.xstate_bv;
|
2008-07-29 17:29:25 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* restore the state passed by the user.
|
|
|
|
*/
|
2008-07-30 00:23:16 +00:00
|
|
|
err = xrestore_user(buf, mask);
|
2008-07-29 17:29:25 +00:00
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* init the state skipped by the user.
|
|
|
|
*/
|
2008-07-30 00:23:16 +00:00
|
|
|
mask = pcntxt_mask & ~mask;
|
2010-06-22 23:23:37 +00:00
|
|
|
if (unlikely(mask))
|
|
|
|
xrstor_state(init_xstate_buf, mask);
|
2008-07-29 17:29:25 +00:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
fx_only:
|
|
|
|
/*
|
|
|
|
* couldn't find the extended state information in the
|
|
|
|
* memory layout. Restore just the FP/SSE and init all
|
|
|
|
* the other extended state.
|
|
|
|
*/
|
2008-07-30 00:23:16 +00:00
|
|
|
xrstor_state(init_xstate_buf, pcntxt_mask & ~XSTATE_FPSSE);
|
2008-07-29 17:29:25 +00:00
|
|
|
return fxrstor_checking((__force struct i387_fxsave_struct *)buf);
|
|
|
|
}
|
|
|
|
|
2008-07-29 17:29:22 +00:00
|
|
|
/*
|
|
|
|
* This restores directly out of user space. Exceptions are handled.
|
|
|
|
*/
|
|
|
|
int restore_i387_xstate(void __user *buf)
|
|
|
|
{
|
|
|
|
struct task_struct *tsk = current;
|
2008-07-29 17:29:25 +00:00
|
|
|
int err = 0;
|
2008-07-29 17:29:22 +00:00
|
|
|
|
|
|
|
if (!buf) {
|
2008-07-29 17:29:25 +00:00
|
|
|
if (used_math())
|
|
|
|
goto clear;
|
2008-07-29 17:29:22 +00:00
|
|
|
return 0;
|
|
|
|
} else
|
|
|
|
if (!access_ok(VERIFY_READ, buf, sig_xstate_size))
|
|
|
|
return -EACCES;
|
|
|
|
|
|
|
|
if (!used_math()) {
|
|
|
|
err = init_fpu(tsk);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
i387: fix x86-64 preemption-unsafe user stack save/restore
Commit 5b1cbac37798 ("i387: make irq_fpu_usable() tests more robust")
added a sanity check to the #NM handler to verify that we never cause
the "Device Not Available" exception in kernel mode.
However, that check actually pinpointed a (fundamental) race where we do
cause that exception as part of the signal stack FPU state save/restore
code.
Because we use the floating point instructions themselves to save and
restore state directly from user mode, we cannot do that atomically with
testing the TS_USEDFPU bit: the user mode access itself may cause a page
fault, which causes a task switch, which saves and restores the FP/MMX
state from the kernel buffers.
This kind of "recursive" FP state save is fine per se, but it means that
when the signal stack save/restore gets restarted, it will now take the
'#NM' exception we originally tried to avoid. With preemption this can
happen even without the page fault - but because of the user access, we
cannot just disable preemption around the save/restore instruction.
There are various ways to solve this, including using the
"enable/disable_page_fault()" helpers to not allow page faults at all
during the sequence, and fall back to copying things by hand without the
use of the native FP state save/restore instructions.
However, the simplest thing to do is to just allow the #NM from kernel
space, but fix the race in setting and clearing CR0.TS that this all
exposed: the TS bit changes and the TS_USEDFPU bit absolutely have to be
atomic wrt scheduling, so while the actual state save/restore can be
interrupted and restarted, the act of actually clearing/setting CR0.TS
and the TS_USEDFPU bit together must not.
Instead of just adding random "preempt_disable/enable()" calls to what
is already excessively ugly code, this introduces some helper functions
that mostly mirror the "kernel_fpu_begin/end()" functionality, just for
the user state instead.
Those helper functions should probably eventually replace the other
ad-hoc CR0.TS and TS_USEDFPU tests too, but I'll need to think about it
some more: the task switching functionality in particular needs to
expose the difference between the 'prev' and 'next' threads, while the
new helper functions intentionally were written to only work with
'current'.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-02-16 17:15:04 +00:00
|
|
|
user_fpu_begin();
|
2010-05-06 08:45:45 +00:00
|
|
|
if (use_xsave())
|
2008-07-29 17:29:25 +00:00
|
|
|
err = restore_user_xstate(buf);
|
|
|
|
else
|
|
|
|
err = fxrstor_checking((__force struct i387_fxsave_struct *)
|
|
|
|
buf);
|
2008-07-29 17:29:22 +00:00
|
|
|
if (unlikely(err)) {
|
|
|
|
/*
|
|
|
|
* Encountered an error while doing the restore from the
|
|
|
|
* user buffer, clear the fpu state.
|
|
|
|
*/
|
2008-07-29 17:29:25 +00:00
|
|
|
clear:
|
2008-07-29 17:29:22 +00:00
|
|
|
clear_fpu(tsk);
|
|
|
|
clear_used_math();
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2008-07-29 17:29:25 +00:00
|
|
|
/*
|
|
|
|
* Prepare the SW reserved portion of the fxsave memory layout, indicating
|
|
|
|
* the presence of the extended state information in the memory layout
|
|
|
|
* pointed by the fpstate pointer in the sigcontext.
|
|
|
|
* This will be saved when ever the FP and extended state context is
|
|
|
|
* saved on the user stack during the signal handler delivery to the user.
|
|
|
|
*/
|
2008-10-21 23:49:09 +00:00
|
|
|
static void prepare_fx_sw_frame(void)
|
2008-07-29 17:29:25 +00:00
|
|
|
{
|
|
|
|
int size_extended = (xstate_size - sizeof(struct i387_fxsave_struct)) +
|
|
|
|
FP_XSTATE_MAGIC2_SIZE;
|
|
|
|
|
|
|
|
sig_xstate_size = sizeof(struct _fpstate) + size_extended;
|
|
|
|
|
|
|
|
#ifdef CONFIG_IA32_EMULATION
|
|
|
|
sig_xstate_ia32_size = sizeof(struct _fpstate_ia32) + size_extended;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
memset(&fx_sw_reserved, 0, sizeof(fx_sw_reserved));
|
|
|
|
|
|
|
|
fx_sw_reserved.magic1 = FP_XSTATE_MAGIC1;
|
|
|
|
fx_sw_reserved.extended_size = sig_xstate_size;
|
2008-07-30 00:23:16 +00:00
|
|
|
fx_sw_reserved.xstate_bv = pcntxt_mask;
|
2008-07-29 17:29:25 +00:00
|
|
|
fx_sw_reserved.xstate_size = xstate_size;
|
|
|
|
#ifdef CONFIG_IA32_EMULATION
|
|
|
|
memcpy(&fx_sw_reserved_ia32, &fx_sw_reserved,
|
|
|
|
sizeof(struct _fpx_sw_bytes));
|
|
|
|
fx_sw_reserved_ia32.extended_size = sig_xstate_ia32_size;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2008-07-29 17:29:21 +00:00
|
|
|
#ifdef CONFIG_X86_64
|
|
|
|
unsigned int sig_xstate_size = sizeof(struct _fpstate);
|
|
|
|
#endif
|
|
|
|
|
2008-07-29 17:29:19 +00:00
|
|
|
/*
|
|
|
|
* Enable the extended processor state save/restore feature
|
|
|
|
*/
|
2010-07-21 21:23:10 +00:00
|
|
|
static inline void xstate_enable(void)
|
2008-07-29 17:29:19 +00:00
|
|
|
{
|
|
|
|
set_in_cr4(X86_CR4_OSXSAVE);
|
2008-07-30 00:23:16 +00:00
|
|
|
xsetbv(XCR_XFEATURE_ENABLED_MASK, pcntxt_mask);
|
2008-07-29 17:29:19 +00:00
|
|
|
}
|
|
|
|
|
2010-07-19 23:05:48 +00:00
|
|
|
/*
|
|
|
|
* Record the offsets and sizes of different state managed by the xsave
|
|
|
|
* memory layout.
|
|
|
|
*/
|
2010-07-21 17:03:56 +00:00
|
|
|
static void __init setup_xstate_features(void)
|
2010-07-19 23:05:48 +00:00
|
|
|
{
|
|
|
|
int eax, ebx, ecx, edx, leaf = 0x2;
|
|
|
|
|
|
|
|
xstate_features = fls64(pcntxt_mask);
|
|
|
|
xstate_offsets = alloc_bootmem(xstate_features * sizeof(int));
|
|
|
|
xstate_sizes = alloc_bootmem(xstate_features * sizeof(int));
|
|
|
|
|
|
|
|
do {
|
2010-07-21 17:03:54 +00:00
|
|
|
cpuid_count(XSTATE_CPUID, leaf, &eax, &ebx, &ecx, &edx);
|
2010-07-19 23:05:48 +00:00
|
|
|
|
|
|
|
if (eax == 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
xstate_offsets[leaf] = ebx;
|
|
|
|
xstate_sizes[leaf] = eax;
|
|
|
|
|
|
|
|
leaf++;
|
|
|
|
} while (1);
|
|
|
|
}
|
|
|
|
|
2008-07-29 17:29:19 +00:00
|
|
|
/*
|
|
|
|
* setup the xstate image representing the init state
|
|
|
|
*/
|
2008-08-30 02:03:34 +00:00
|
|
|
static void __init setup_xstate_init(void)
|
2008-07-29 17:29:19 +00:00
|
|
|
{
|
2010-07-19 23:05:49 +00:00
|
|
|
setup_xstate_features();
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Setup init_xstate_buf to represent the init state of
|
|
|
|
* all the features managed by the xsave
|
|
|
|
*/
|
2010-11-16 21:23:51 +00:00
|
|
|
init_xstate_buf = alloc_bootmem_align(xstate_size,
|
|
|
|
__alignof__(struct xsave_struct));
|
2008-07-29 17:29:19 +00:00
|
|
|
init_xstate_buf->i387.mxcsr = MXCSR_DEFAULT;
|
2010-07-19 23:05:48 +00:00
|
|
|
|
2010-07-19 23:05:49 +00:00
|
|
|
clts();
|
|
|
|
/*
|
|
|
|
* Init all the features state with header_bv being 0x0
|
|
|
|
*/
|
|
|
|
xrstor_state(init_xstate_buf, -1);
|
|
|
|
/*
|
|
|
|
* Dump the init state again. This is to identify the init state
|
|
|
|
* of any feature which is not represented by all zero's.
|
|
|
|
*/
|
|
|
|
xsave_state(init_xstate_buf, -1);
|
|
|
|
stts();
|
2008-07-29 17:29:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Enable and initialize the xsave feature.
|
|
|
|
*/
|
2010-07-21 21:23:10 +00:00
|
|
|
static void __init xstate_enable_boot_cpu(void)
|
2008-07-29 17:29:19 +00:00
|
|
|
{
|
|
|
|
unsigned int eax, ebx, ecx, edx;
|
|
|
|
|
2010-07-21 17:03:54 +00:00
|
|
|
if (boot_cpu_data.cpuid_level < XSTATE_CPUID) {
|
|
|
|
WARN(1, KERN_ERR "XSTATE_CPUID missing\n");
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
cpuid_count(XSTATE_CPUID, 0, &eax, &ebx, &ecx, &edx);
|
2008-07-30 00:23:16 +00:00
|
|
|
pcntxt_mask = eax + ((u64)edx << 32);
|
2008-07-29 17:29:19 +00:00
|
|
|
|
2008-07-30 00:23:16 +00:00
|
|
|
if ((pcntxt_mask & XSTATE_FPSSE) != XSTATE_FPSSE) {
|
|
|
|
printk(KERN_ERR "FP/SSE not shown under xsave features 0x%llx\n",
|
|
|
|
pcntxt_mask);
|
2008-07-29 17:29:19 +00:00
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2009-04-10 22:21:24 +00:00
|
|
|
* Support only the state known to OS.
|
2008-07-29 17:29:19 +00:00
|
|
|
*/
|
2008-07-30 00:23:16 +00:00
|
|
|
pcntxt_mask = pcntxt_mask & XCNTXT_MASK;
|
2010-07-21 17:03:53 +00:00
|
|
|
|
2010-07-21 21:23:10 +00:00
|
|
|
xstate_enable();
|
2008-07-29 17:29:19 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Recompute the context size for enabled features
|
|
|
|
*/
|
2010-07-21 17:03:54 +00:00
|
|
|
cpuid_count(XSTATE_CPUID, 0, &eax, &ebx, &ecx, &edx);
|
2008-07-29 17:29:19 +00:00
|
|
|
xstate_size = ebx;
|
|
|
|
|
2010-02-11 19:50:59 +00:00
|
|
|
update_regset_xstate_info(xstate_size, pcntxt_mask);
|
2008-07-29 17:29:25 +00:00
|
|
|
prepare_fx_sw_frame();
|
|
|
|
|
2008-07-29 17:29:19 +00:00
|
|
|
setup_xstate_init();
|
|
|
|
|
2008-07-30 00:23:16 +00:00
|
|
|
printk(KERN_INFO "xsave/xrstor: enabled xstate_bv 0x%llx, "
|
2008-07-29 17:29:19 +00:00
|
|
|
"cntxt size 0x%x\n",
|
2008-07-30 00:23:16 +00:00
|
|
|
pcntxt_mask, xstate_size);
|
2008-07-29 17:29:19 +00:00
|
|
|
}
|
2010-07-20 18:50:51 +00:00
|
|
|
|
2010-07-21 21:23:10 +00:00
|
|
|
/*
|
|
|
|
* For the very first instance, this calls xstate_enable_boot_cpu();
|
|
|
|
* for all subsequent instances, this calls xstate_enable().
|
|
|
|
*
|
|
|
|
* This is somewhat obfuscated due to the lack of powerful enough
|
|
|
|
* overrides for the section checks.
|
|
|
|
*/
|
2010-07-20 18:50:51 +00:00
|
|
|
void __cpuinit xsave_init(void)
|
|
|
|
{
|
2010-07-21 21:23:10 +00:00
|
|
|
static __refdata void (*next_func)(void) = xstate_enable_boot_cpu;
|
|
|
|
void (*this_func)(void);
|
|
|
|
|
2010-07-21 17:03:52 +00:00
|
|
|
if (!cpu_has_xsave)
|
|
|
|
return;
|
|
|
|
|
2010-07-21 21:23:10 +00:00
|
|
|
this_func = next_func;
|
|
|
|
next_func = xstate_enable;
|
|
|
|
this_func();
|
2010-07-20 18:50:51 +00:00
|
|
|
}
|