License cleanup: add SPDX GPL-2.0 license identifier to files with no license
Many source files in the tree are missing licensing information, which
makes it harder for compliance tools to determine the correct license.
By default all files without license information are under the default
license of the kernel, which is GPL version 2.
Update the files which contain no license information with the 'GPL-2.0'
SPDX license identifier. The SPDX identifier is a legally binding
shorthand, which can be used instead of the full boiler plate text.
This patch is based on work done by Thomas Gleixner and Kate Stewart and
Philippe Ombredanne.
How this work was done:
Patches were generated and checked against linux-4.14-rc6 for a subset of
the use cases:
- file had no licensing information it it.
- file was a */uapi/* one with no licensing information in it,
- file was a */uapi/* one with existing licensing information,
Further patches will be generated in subsequent months to fix up cases
where non-standard license headers were used, and references to license
had to be inferred by heuristics based on keywords.
The analysis to determine which SPDX License Identifier to be applied to
a file was done in a spreadsheet of side by side results from of the
output of two independent scanners (ScanCode & Windriver) producing SPDX
tag:value files created by Philippe Ombredanne. Philippe prepared the
base worksheet, and did an initial spot review of a few 1000 files.
The 4.13 kernel was the starting point of the analysis with 60,537 files
assessed. Kate Stewart did a file by file comparison of the scanner
results in the spreadsheet to determine which SPDX license identifier(s)
to be applied to the file. She confirmed any determination that was not
immediately clear with lawyers working with the Linux Foundation.
Criteria used to select files for SPDX license identifier tagging was:
- Files considered eligible had to be source code files.
- Make and config files were included as candidates if they contained >5
lines of source
- File already had some variant of a license header in it (even if <5
lines).
All documentation files were explicitly excluded.
The following heuristics were used to determine which SPDX license
identifiers to apply.
- when both scanners couldn't find any license traces, file was
considered to have no license information in it, and the top level
COPYING file license applied.
For non */uapi/* files that summary was:
SPDX license identifier # files
---------------------------------------------------|-------
GPL-2.0 11139
and resulted in the first patch in this series.
If that file was a */uapi/* path one, it was "GPL-2.0 WITH
Linux-syscall-note" otherwise it was "GPL-2.0". Results of that was:
SPDX license identifier # files
---------------------------------------------------|-------
GPL-2.0 WITH Linux-syscall-note 930
and resulted in the second patch in this series.
- if a file had some form of licensing information in it, and was one
of the */uapi/* ones, it was denoted with the Linux-syscall-note if
any GPL family license was found in the file or had no licensing in
it (per prior point). Results summary:
SPDX license identifier # files
---------------------------------------------------|------
GPL-2.0 WITH Linux-syscall-note 270
GPL-2.0+ WITH Linux-syscall-note 169
((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) 21
((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause) 17
LGPL-2.1+ WITH Linux-syscall-note 15
GPL-1.0+ WITH Linux-syscall-note 14
((GPL-2.0+ WITH Linux-syscall-note) OR BSD-3-Clause) 5
LGPL-2.0+ WITH Linux-syscall-note 4
LGPL-2.1 WITH Linux-syscall-note 3
((GPL-2.0 WITH Linux-syscall-note) OR MIT) 3
((GPL-2.0 WITH Linux-syscall-note) AND MIT) 1
and that resulted in the third patch in this series.
- when the two scanners agreed on the detected license(s), that became
the concluded license(s).
- when there was disagreement between the two scanners (one detected a
license but the other didn't, or they both detected different
licenses) a manual inspection of the file occurred.
- In most cases a manual inspection of the information in the file
resulted in a clear resolution of the license that should apply (and
which scanner probably needed to revisit its heuristics).
- When it was not immediately clear, the license identifier was
confirmed with lawyers working with the Linux Foundation.
- If there was any question as to the appropriate license identifier,
the file was flagged for further research and to be revisited later
in time.
In total, over 70 hours of logged manual review was done on the
spreadsheet to determine the SPDX license identifiers to apply to the
source files by Kate, Philippe, Thomas and, in some cases, confirmation
by lawyers working with the Linux Foundation.
Kate also obtained a third independent scan of the 4.13 code base from
FOSSology, and compared selected files where the other two scanners
disagreed against that SPDX file, to see if there was new insights. The
Windriver scanner is based on an older version of FOSSology in part, so
they are related.
Thomas did random spot checks in about 500 files from the spreadsheets
for the uapi headers and agreed with SPDX license identifier in the
files he inspected. For the non-uapi files Thomas did random spot checks
in about 15000 files.
In initial set of patches against 4.14-rc6, 3 files were found to have
copy/paste license identifier errors, and have been fixed to reflect the
correct identifier.
Additionally Philippe spent 10 hours this week doing a detailed manual
inspection and review of the 12,461 patched files from the initial patch
version early this week with:
- a full scancode scan run, collecting the matched texts, detected
license ids and scores
- reviewing anything where there was a license detected (about 500+
files) to ensure that the applied SPDX license was correct
- reviewing anything where there was no detection but the patch license
was not GPL-2.0 WITH Linux-syscall-note to ensure that the applied
SPDX license was correct
This produced a worksheet with 20 files needing minor correction. This
worksheet was then exported into 3 different .csv files for the
different types of files to be modified.
These .csv files were then reviewed by Greg. Thomas wrote a script to
parse the csv files and add the proper SPDX tag to the file, in the
format that the file expected. This script was further refined by Greg
based on the output to detect more types of files automatically and to
distinguish between header and source .c files (which need different
comment types.) Finally Greg ran the script using the .csv files to
generate the patches.
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Reviewed-by: Philippe Ombredanne <pombredanne@nexb.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-11-01 14:07:57 +00:00
|
|
|
// SPDX-License-Identifier: GPL-2.0
|
2008-05-19 23:52:27 +00:00
|
|
|
/*
|
2005-04-16 22:20:36 +00:00
|
|
|
* unaligned.c: Unaligned load/store trap handling with special
|
|
|
|
|
* cases for the kernel to do them more quickly.
|
|
|
|
|
*
|
2008-07-18 05:11:32 +00:00
|
|
|
* Copyright (C) 1996,2008 David S. Miller (davem@davemloft.net)
|
2005-04-16 22:20:36 +00:00
|
|
|
* Copyright (C) 1996,1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
|
2008-02-18 07:24:10 +00:00
|
|
|
#include <linux/jiffies.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/kernel.h>
|
|
|
|
|
#include <linux/sched.h>
|
|
|
|
|
#include <linux/mm.h>
|
2016-09-19 21:36:29 +00:00
|
|
|
#include <linux/extable.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <asm/asi.h>
|
|
|
|
|
#include <asm/ptrace.h>
|
|
|
|
|
#include <asm/pstate.h>
|
|
|
|
|
#include <asm/processor.h>
|
2016-12-24 19:46:01 +00:00
|
|
|
#include <linux/uaccess.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/smp.h>
|
|
|
|
|
#include <linux/bitops.h>
|
2009-12-11 09:07:53 +00:00
|
|
|
#include <linux/perf_event.h>
|
2010-02-28 11:31:29 +00:00
|
|
|
#include <linux/ratelimit.h>
|
2013-09-14 12:02:11 +00:00
|
|
|
#include <linux/context_tracking.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <asm/fpumacro.h>
|
2012-03-28 17:30:03 +00:00
|
|
|
#include <asm/cacheflush.h>
|
2014-05-16 21:25:54 +00:00
|
|
|
#include <asm/setup.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-09-14 12:02:11 +00:00
|
|
|
#include "entry.h"
|
2014-05-16 21:25:54 +00:00
|
|
|
#include "kernel.h"
|
2013-09-14 12:02:11 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
enum direction {
|
|
|
|
|
load, /* ld, ldd, ldh, ldsh */
|
|
|
|
|
store, /* st, std, sth, stsh */
|
|
|
|
|
both, /* Swap, ldstub, cas, ... */
|
|
|
|
|
fpld,
|
|
|
|
|
fpst,
|
|
|
|
|
invalid,
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
static inline enum direction decode_direction(unsigned int insn)
|
|
|
|
|
{
|
|
|
|
|
unsigned long tmp = (insn >> 21) & 1;
|
|
|
|
|
|
|
|
|
|
if (!tmp)
|
|
|
|
|
return load;
|
|
|
|
|
else {
|
|
|
|
|
switch ((insn>>19)&0xf) {
|
|
|
|
|
case 15: /* swap* */
|
|
|
|
|
return both;
|
|
|
|
|
default:
|
|
|
|
|
return store;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* 16 = double-word, 8 = extra-word, 4 = word, 2 = half-word */
|
2010-04-19 20:46:48 +00:00
|
|
|
static inline int decode_access_size(struct pt_regs *regs, unsigned int insn)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
|
unsigned int tmp;
|
|
|
|
|
|
|
|
|
|
tmp = ((insn >> 19) & 0xf);
|
|
|
|
|
if (tmp == 11 || tmp == 14) /* ldx/stx */
|
|
|
|
|
return 8;
|
|
|
|
|
tmp &= 3;
|
|
|
|
|
if (!tmp)
|
|
|
|
|
return 4;
|
|
|
|
|
else if (tmp == 3)
|
|
|
|
|
return 16; /* ldd/std - Although it is actually 8 */
|
|
|
|
|
else if (tmp == 2)
|
|
|
|
|
return 2;
|
|
|
|
|
else {
|
|
|
|
|
printk("Impossible unaligned trap. insn=%08x\n", insn);
|
2010-04-19 20:46:48 +00:00
|
|
|
die_if_kernel("Byte sized unaligned access?!?!", regs);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
|
/* GCC should never warn that control reaches the end
|
|
|
|
|
* of this function without returning a value because
|
|
|
|
|
* die_if_kernel() is marked with attribute 'noreturn'.
|
|
|
|
|
* Alas, some versions do...
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static inline int decode_asi(unsigned int insn, struct pt_regs *regs)
|
|
|
|
|
{
|
|
|
|
|
if (insn & 0x800000) {
|
|
|
|
|
if (insn & 0x2000)
|
|
|
|
|
return (unsigned char)(regs->tstate >> 24); /* %asi */
|
|
|
|
|
else
|
|
|
|
|
return (unsigned char)(insn >> 5); /* imm_asi */
|
|
|
|
|
} else
|
|
|
|
|
return ASI_P;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* 0x400000 = signed, 0 = unsigned */
|
|
|
|
|
static inline int decode_signedness(unsigned int insn)
|
|
|
|
|
{
|
|
|
|
|
return (insn & 0x400000);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static inline void maybe_flush_windows(unsigned int rs1, unsigned int rs2,
|
|
|
|
|
unsigned int rd, int from_kernel)
|
|
|
|
|
{
|
|
|
|
|
if (rs2 >= 16 || rs1 >= 16 || rd >= 16) {
|
|
|
|
|
if (from_kernel != 0)
|
|
|
|
|
__asm__ __volatile__("flushw");
|
|
|
|
|
else
|
|
|
|
|
flushw_user();
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static inline long sign_extend_imm13(long imm)
|
|
|
|
|
{
|
|
|
|
|
return imm << 51 >> 51;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
|
|
|
|
|
{
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
unsigned long value, fp;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
|
if (reg < 16)
|
|
|
|
|
return (!reg ? 0 : regs->u_regs[reg]);
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
|
|
|
|
|
fp = regs->u_regs[UREG_FP];
|
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if (regs->tstate & TSTATE_PRIV) {
|
|
|
|
|
struct reg_window *win;
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
win = (struct reg_window *)(fp + STACK_BIAS);
|
2005-04-16 22:20:36 +00:00
|
|
|
value = win->locals[reg - 16];
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
} else if (!test_thread_64bit_stack(fp)) {
|
2005-04-16 22:20:36 +00:00
|
|
|
struct reg_window32 __user *win32;
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
|
2005-04-16 22:20:36 +00:00
|
|
|
get_user(value, &win32->locals[reg - 16]);
|
|
|
|
|
} else {
|
|
|
|
|
struct reg_window __user *win;
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
win = (struct reg_window __user *)(fp + STACK_BIAS);
|
2005-04-16 22:20:36 +00:00
|
|
|
get_user(value, &win->locals[reg - 16]);
|
|
|
|
|
}
|
|
|
|
|
return value;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static unsigned long *fetch_reg_addr(unsigned int reg, struct pt_regs *regs)
|
|
|
|
|
{
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
unsigned long fp;
|
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if (reg < 16)
|
|
|
|
|
return ®s->u_regs[reg];
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
|
|
|
|
|
fp = regs->u_regs[UREG_FP];
|
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if (regs->tstate & TSTATE_PRIV) {
|
|
|
|
|
struct reg_window *win;
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
win = (struct reg_window *)(fp + STACK_BIAS);
|
2005-04-16 22:20:36 +00:00
|
|
|
return &win->locals[reg - 16];
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
} else if (!test_thread_64bit_stack(fp)) {
|
2005-04-16 22:20:36 +00:00
|
|
|
struct reg_window32 *win32;
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
win32 = (struct reg_window32 *)((unsigned long)((u32)fp));
|
2005-04-16 22:20:36 +00:00
|
|
|
return (unsigned long *)&win32->locals[reg - 16];
|
|
|
|
|
} else {
|
|
|
|
|
struct reg_window *win;
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
win = (struct reg_window *)(fp + STACK_BIAS);
|
2005-04-16 22:20:36 +00:00
|
|
|
return &win->locals[reg - 16];
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
unsigned long compute_effective_address(struct pt_regs *regs,
|
|
|
|
|
unsigned int insn, unsigned int rd)
|
|
|
|
|
{
|
2014-04-29 06:50:08 +00:00
|
|
|
int from_kernel = (regs->tstate & TSTATE_PRIV) != 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned int rs1 = (insn >> 14) & 0x1f;
|
|
|
|
|
unsigned int rs2 = insn & 0x1f;
|
2014-04-29 06:50:08 +00:00
|
|
|
unsigned long addr;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
|
if (insn & 0x2000) {
|
|
|
|
|
maybe_flush_windows(rs1, 0, rd, from_kernel);
|
2014-04-29 06:50:08 +00:00
|
|
|
addr = (fetch_reg(rs1, regs) + sign_extend_imm13(insn));
|
2005-04-16 22:20:36 +00:00
|
|
|
} else {
|
|
|
|
|
maybe_flush_windows(rs1, rs2, rd, from_kernel);
|
2014-04-29 06:50:08 +00:00
|
|
|
addr = (fetch_reg(rs1, regs) + fetch_reg(rs2, regs));
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2014-04-29 06:50:08 +00:00
|
|
|
|
|
|
|
|
if (!from_kernel && test_thread_flag(TIF_32BIT))
|
|
|
|
|
addr &= 0xffffffff;
|
|
|
|
|
|
|
|
|
|
return addr;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* This is just to make gcc think die_if_kernel does return... */
|
2008-01-24 21:16:20 +00:00
|
|
|
static void __used unaligned_panic(char *str, struct pt_regs *regs)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
|
die_if_kernel(str, regs);
|
|
|
|
|
}
|
|
|
|
|
|
2005-09-29 03:41:45 +00:00
|
|
|
extern int do_int_load(unsigned long *dest_reg, int size,
|
|
|
|
|
unsigned long *saddr, int is_signed, int asi);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2005-09-29 03:41:45 +00:00
|
|
|
extern int __do_int_store(unsigned long *dst_addr, int size,
|
|
|
|
|
unsigned long src_val, int asi);
|
2005-08-19 22:55:33 +00:00
|
|
|
|
2005-09-29 03:41:45 +00:00
|
|
|
static inline int do_int_store(int reg_num, int size, unsigned long *dst_addr,
|
|
|
|
|
struct pt_regs *regs, int asi, int orig_asi)
|
2005-08-19 22:55:33 +00:00
|
|
|
{
|
|
|
|
|
unsigned long zero = 0;
|
2005-09-20 02:56:06 +00:00
|
|
|
unsigned long *src_val_p = &zero;
|
|
|
|
|
unsigned long src_val;
|
2005-08-19 22:55:33 +00:00
|
|
|
|
|
|
|
|
if (size == 16) {
|
|
|
|
|
size = 8;
|
|
|
|
|
zero = (((long)(reg_num ?
|
2016-03-10 23:21:43 +00:00
|
|
|
(unsigned int)fetch_reg(reg_num, regs) : 0)) << 32) |
|
|
|
|
|
(unsigned int)fetch_reg(reg_num + 1, regs);
|
2005-08-19 22:55:33 +00:00
|
|
|
} else if (reg_num) {
|
2005-09-20 02:56:06 +00:00
|
|
|
src_val_p = fetch_reg_addr(reg_num, regs);
|
|
|
|
|
}
|
|
|
|
|
src_val = *src_val_p;
|
|
|
|
|
if (unlikely(asi != orig_asi)) {
|
|
|
|
|
switch (size) {
|
|
|
|
|
case 2:
|
|
|
|
|
src_val = swab16(src_val);
|
|
|
|
|
break;
|
|
|
|
|
case 4:
|
|
|
|
|
src_val = swab32(src_val);
|
|
|
|
|
break;
|
|
|
|
|
case 8:
|
|
|
|
|
src_val = swab64(src_val);
|
|
|
|
|
break;
|
|
|
|
|
case 16:
|
|
|
|
|
default:
|
|
|
|
|
BUG();
|
|
|
|
|
break;
|
2011-06-03 14:45:23 +00:00
|
|
|
}
|
2005-08-19 22:55:33 +00:00
|
|
|
}
|
2005-09-29 03:41:45 +00:00
|
|
|
return __do_int_store(dst_addr, size, src_val, asi);
|
2005-08-19 22:55:33 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
|
static inline void advance(struct pt_regs *regs)
|
|
|
|
|
{
|
|
|
|
|
regs->tpc = regs->tnpc;
|
|
|
|
|
regs->tnpc += 4;
|
|
|
|
|
if (test_thread_flag(TIF_32BIT)) {
|
|
|
|
|
regs->tpc &= 0xffffffff;
|
|
|
|
|
regs->tnpc &= 0xffffffff;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static inline int floating_point_load_or_store_p(unsigned int insn)
|
|
|
|
|
{
|
|
|
|
|
return (insn >> 24) & 1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static inline int ok_for_kernel(unsigned int insn)
|
|
|
|
|
{
|
|
|
|
|
return !floating_point_load_or_store_p(insn);
|
|
|
|
|
}
|
|
|
|
|
|
2006-11-29 04:18:05 +00:00
|
|
|
static void kernel_mna_trap_fault(int fixup_tstate_asi)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2005-08-19 22:55:33 +00:00
|
|
|
struct pt_regs *regs = current_thread_info()->kern_una_regs;
|
|
|
|
|
unsigned int insn = current_thread_info()->kern_una_insn;
|
2005-09-29 03:21:11 +00:00
|
|
|
const struct exception_table_entry *entry;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2005-09-29 03:21:11 +00:00
|
|
|
entry = search_exception_tables(regs->tpc);
|
|
|
|
|
if (!entry) {
|
2005-08-19 22:55:33 +00:00
|
|
|
unsigned long address;
|
|
|
|
|
|
|
|
|
|
address = compute_effective_address(regs, insn,
|
|
|
|
|
((insn >> 25) & 0x1f));
|
2005-04-16 22:20:36 +00:00
|
|
|
if (address < PAGE_SIZE) {
|
2005-08-19 22:55:33 +00:00
|
|
|
printk(KERN_ALERT "Unable to handle kernel NULL "
|
|
|
|
|
"pointer dereference in mna handler");
|
2005-04-16 22:20:36 +00:00
|
|
|
} else
|
2005-08-19 22:55:33 +00:00
|
|
|
printk(KERN_ALERT "Unable to handle kernel paging "
|
|
|
|
|
"request in mna handler");
|
2005-04-16 22:20:36 +00:00
|
|
|
printk(KERN_ALERT " at virtual address %016lx\n",address);
|
2005-08-19 22:55:33 +00:00
|
|
|
printk(KERN_ALERT "current->{active_,}mm->context = %016lx\n",
|
2005-04-16 22:20:36 +00:00
|
|
|
(current->mm ? CTX_HWBITS(current->mm->context) :
|
|
|
|
|
CTX_HWBITS(current->active_mm->context)));
|
2005-08-19 22:55:33 +00:00
|
|
|
printk(KERN_ALERT "current->{active_,}mm->pgd = %016lx\n",
|
2005-04-16 22:20:36 +00:00
|
|
|
(current->mm ? (unsigned long) current->mm->pgd :
|
|
|
|
|
(unsigned long) current->active_mm->pgd));
|
|
|
|
|
die_if_kernel("Oops", regs);
|
|
|
|
|
/* Not reached */
|
|
|
|
|
}
|
2005-09-29 03:21:11 +00:00
|
|
|
regs->tpc = entry->fixup;
|
2005-04-16 22:20:36 +00:00
|
|
|
regs->tnpc = regs->tpc + 4;
|
|
|
|
|
|
2006-11-29 04:18:05 +00:00
|
|
|
if (fixup_tstate_asi) {
|
|
|
|
|
regs->tstate &= ~TSTATE_ASI;
|
|
|
|
|
regs->tstate |= (ASI_AIUS << 24UL);
|
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
2006-11-29 04:18:05 +00:00
|
|
|
static void log_unaligned(struct pt_regs *regs)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2010-02-28 11:31:29 +00:00
|
|
|
static DEFINE_RATELIMIT_STATE(ratelimit, 5 * HZ, 5);
|
2005-08-19 22:55:33 +00:00
|
|
|
|
2010-02-28 11:31:29 +00:00
|
|
|
if (__ratelimit(&ratelimit)) {
|
2008-07-18 05:11:32 +00:00
|
|
|
printk("Kernel unaligned access at TPC[%lx] %pS\n",
|
|
|
|
|
regs->tpc, (void *) regs->tpc);
|
2006-06-22 05:31:08 +00:00
|
|
|
}
|
2006-11-29 04:18:05 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
asmlinkage void kernel_unaligned_trap(struct pt_regs *regs, unsigned int insn)
|
|
|
|
|
{
|
|
|
|
|
enum direction dir = decode_direction(insn);
|
2010-04-19 20:46:48 +00:00
|
|
|
int size = decode_access_size(regs, insn);
|
2006-11-29 04:18:05 +00:00
|
|
|
int orig_asi, asi;
|
|
|
|
|
|
|
|
|
|
current_thread_info()->kern_una_regs = regs;
|
|
|
|
|
current_thread_info()->kern_una_insn = insn;
|
|
|
|
|
|
|
|
|
|
orig_asi = asi = decode_asi(insn, regs);
|
|
|
|
|
|
|
|
|
|
/* If this is a {get,put}_user() on an unaligned userspace pointer,
|
|
|
|
|
* just signal a fault and do not log the event.
|
|
|
|
|
*/
|
|
|
|
|
if (asi == ASI_AIUS) {
|
|
|
|
|
kernel_mna_trap_fault(0);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
log_unaligned(regs);
|
2006-06-22 05:31:08 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if (!ok_for_kernel(insn) || dir == both) {
|
2005-08-19 22:55:33 +00:00
|
|
|
printk("Unsupported unaligned load/store trap for kernel "
|
|
|
|
|
"at <%016lx>.\n", regs->tpc);
|
|
|
|
|
unaligned_panic("Kernel does fpu/atomic "
|
|
|
|
|
"unaligned load/store.", regs);
|
|
|
|
|
|
2006-11-29 04:18:05 +00:00
|
|
|
kernel_mna_trap_fault(0);
|
2005-04-16 22:20:36 +00:00
|
|
|
} else {
|
2005-09-28 23:48:40 +00:00
|
|
|
unsigned long addr, *reg_addr;
|
2006-11-29 04:18:05 +00:00
|
|
|
int err;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2005-08-19 22:55:33 +00:00
|
|
|
addr = compute_effective_address(regs, insn,
|
|
|
|
|
((insn >> 25) & 0x1f));
|
2011-06-27 12:41:57 +00:00
|
|
|
perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, addr);
|
2005-09-20 02:56:06 +00:00
|
|
|
switch (asi) {
|
|
|
|
|
case ASI_NL:
|
|
|
|
|
case ASI_AIUPL:
|
|
|
|
|
case ASI_AIUSL:
|
|
|
|
|
case ASI_PL:
|
|
|
|
|
case ASI_SL:
|
|
|
|
|
case ASI_PNFL:
|
|
|
|
|
case ASI_SNFL:
|
|
|
|
|
asi &= ~0x08;
|
|
|
|
|
break;
|
2011-06-03 14:45:23 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
switch (dir) {
|
|
|
|
|
case load:
|
2005-09-28 23:48:40 +00:00
|
|
|
reg_addr = fetch_reg_addr(((insn>>25)&0x1f), regs);
|
2005-09-29 03:41:45 +00:00
|
|
|
err = do_int_load(reg_addr, size,
|
|
|
|
|
(unsigned long *) addr,
|
|
|
|
|
decode_signedness(insn), asi);
|
|
|
|
|
if (likely(!err) && unlikely(asi != orig_asi)) {
|
2005-09-28 23:48:40 +00:00
|
|
|
unsigned long val_in = *reg_addr;
|
2005-09-20 02:56:06 +00:00
|
|
|
switch (size) {
|
|
|
|
|
case 2:
|
|
|
|
|
val_in = swab16(val_in);
|
|
|
|
|
break;
|
|
|
|
|
case 4:
|
|
|
|
|
val_in = swab32(val_in);
|
|
|
|
|
break;
|
|
|
|
|
case 8:
|
|
|
|
|
val_in = swab64(val_in);
|
|
|
|
|
break;
|
|
|
|
|
case 16:
|
|
|
|
|
default:
|
|
|
|
|
BUG();
|
|
|
|
|
break;
|
2011-06-03 14:45:23 +00:00
|
|
|
}
|
2005-09-28 23:48:40 +00:00
|
|
|
*reg_addr = val_in;
|
2005-09-20 02:56:06 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
case store:
|
2005-09-29 03:41:45 +00:00
|
|
|
err = do_int_store(((insn>>25)&0x1f), size,
|
|
|
|
|
(unsigned long *) addr, regs,
|
|
|
|
|
asi, orig_asi);
|
2005-04-16 22:20:36 +00:00
|
|
|
break;
|
2005-08-19 22:55:33 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
default:
|
|
|
|
|
panic("Impossible kernel unaligned trap.");
|
|
|
|
|
/* Not reached... */
|
|
|
|
|
}
|
2005-09-29 03:41:45 +00:00
|
|
|
if (unlikely(err))
|
2006-11-29 04:18:05 +00:00
|
|
|
kernel_mna_trap_fault(1);
|
2005-09-29 03:41:45 +00:00
|
|
|
else
|
|
|
|
|
advance(regs);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int handle_popc(u32 insn, struct pt_regs *regs)
|
|
|
|
|
{
|
|
|
|
|
int from_kernel = (regs->tstate & TSTATE_PRIV) != 0;
|
2011-08-02 02:41:12 +00:00
|
|
|
int ret, rd = ((insn >> 25) & 0x1f);
|
|
|
|
|
u64 value;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2011-06-27 12:41:57 +00:00
|
|
|
perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (insn & 0x2000) {
|
|
|
|
|
maybe_flush_windows(0, 0, rd, from_kernel);
|
|
|
|
|
value = sign_extend_imm13(insn);
|
|
|
|
|
} else {
|
|
|
|
|
maybe_flush_windows(0, insn & 0x1f, rd, from_kernel);
|
|
|
|
|
value = fetch_reg(insn & 0x1f, regs);
|
|
|
|
|
}
|
2011-08-02 02:41:12 +00:00
|
|
|
ret = hweight64(value);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (rd < 16) {
|
|
|
|
|
if (rd)
|
|
|
|
|
regs->u_regs[rd] = ret;
|
|
|
|
|
} else {
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
unsigned long fp = regs->u_regs[UREG_FP];
|
|
|
|
|
|
|
|
|
|
if (!test_thread_64bit_stack(fp)) {
|
2005-04-16 22:20:36 +00:00
|
|
|
struct reg_window32 __user *win32;
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
|
2005-04-16 22:20:36 +00:00
|
|
|
put_user(ret, &win32->locals[rd - 16]);
|
|
|
|
|
} else {
|
|
|
|
|
struct reg_window __user *win;
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
win = (struct reg_window __user *)(fp + STACK_BIAS);
|
2005-04-16 22:20:36 +00:00
|
|
|
put_user(ret, &win->locals[rd - 16]);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
advance(regs);
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
extern void do_fpother(struct pt_regs *regs);
|
|
|
|
|
extern void do_privact(struct pt_regs *regs);
|
2006-02-10 04:20:34 +00:00
|
|
|
extern void sun4v_data_access_exception(struct pt_regs *regs,
|
|
|
|
|
unsigned long addr,
|
|
|
|
|
unsigned long type_ctx);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
|
int handle_ldf_stq(u32 insn, struct pt_regs *regs)
|
|
|
|
|
{
|
|
|
|
|
unsigned long addr = compute_effective_address(regs, insn, 0);
|
sparc64: Don't restrict fp regs for no-fault loads
The function handle_ldf_stq() deals with no-fault ASI
loads and stores, but restricts fp registers to quad
word regs (ie, %f0, %f4 etc). This is valid for the
STQ case, but unnecessarily restricts loads, which
may be single precision, double, or quad. This results
in SIGFPE being raised for this instruction when the
source address is invalid:
ldda [%g1] ASI_PNF, %f2
but not for this one:
ldda [%g1] ASI_PNF, %f4
The validation check for quad register is moved to
within the STQ block so that loads are not affected
by the check.
An additional problem is that the calculation for freg
is incorrect when a single precision load is being
handled. This causes %f1 to be seen as %f32 etc,
and the incorrect register ends up being overwritten.
This code sequence demonstrates the problem:
ldd [%g1], %f32 ! g1 = valid address
lda [%i3] ASI_PNF, %f1 ! i3 = invalid address
std %f32, [%g1]
This is corrected by basing the freg calculation on
the load size.
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-10-31 04:36:23 +00:00
|
|
|
int freg;
|
2005-04-16 22:20:36 +00:00
|
|
|
struct fpustate *f = FPUSTATE;
|
|
|
|
|
int asi = decode_asi(insn, regs);
|
sparc64: Don't restrict fp regs for no-fault loads
The function handle_ldf_stq() deals with no-fault ASI
loads and stores, but restricts fp registers to quad
word regs (ie, %f0, %f4 etc). This is valid for the
STQ case, but unnecessarily restricts loads, which
may be single precision, double, or quad. This results
in SIGFPE being raised for this instruction when the
source address is invalid:
ldda [%g1] ASI_PNF, %f2
but not for this one:
ldda [%g1] ASI_PNF, %f4
The validation check for quad register is moved to
within the STQ block so that loads are not affected
by the check.
An additional problem is that the calculation for freg
is incorrect when a single precision load is being
handled. This causes %f1 to be seen as %f32 etc,
and the incorrect register ends up being overwritten.
This code sequence demonstrates the problem:
ldd [%g1], %f32 ! g1 = valid address
lda [%i3] ASI_PNF, %f1 ! i3 = invalid address
std %f32, [%g1]
This is corrected by basing the freg calculation on
the load size.
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-10-31 04:36:23 +00:00
|
|
|
int flag;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2011-06-27 12:41:57 +00:00
|
|
|
perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0);
|
2009-12-11 09:07:53 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
save_and_clear_fpu();
|
|
|
|
|
current_thread_info()->xfsr[0] &= ~0x1c000;
|
|
|
|
|
if (insn & 0x200000) {
|
|
|
|
|
/* STQ */
|
|
|
|
|
u64 first = 0, second = 0;
|
|
|
|
|
|
sparc64: Don't restrict fp regs for no-fault loads
The function handle_ldf_stq() deals with no-fault ASI
loads and stores, but restricts fp registers to quad
word regs (ie, %f0, %f4 etc). This is valid for the
STQ case, but unnecessarily restricts loads, which
may be single precision, double, or quad. This results
in SIGFPE being raised for this instruction when the
source address is invalid:
ldda [%g1] ASI_PNF, %f2
but not for this one:
ldda [%g1] ASI_PNF, %f4
The validation check for quad register is moved to
within the STQ block so that loads are not affected
by the check.
An additional problem is that the calculation for freg
is incorrect when a single precision load is being
handled. This causes %f1 to be seen as %f32 etc,
and the incorrect register ends up being overwritten.
This code sequence demonstrates the problem:
ldd [%g1], %f32 ! g1 = valid address
lda [%i3] ASI_PNF, %f1 ! i3 = invalid address
std %f32, [%g1]
This is corrected by basing the freg calculation on
the load size.
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-10-31 04:36:23 +00:00
|
|
|
freg = ((insn >> 25) & 0x1e) | ((insn >> 20) & 0x20);
|
|
|
|
|
flag = (freg < 32) ? FPRS_DL : FPRS_DU;
|
|
|
|
|
if (freg & 3) {
|
|
|
|
|
current_thread_info()->xfsr[0] |= (6 << 14) /* invalid_fp_register */;
|
|
|
|
|
do_fpother(regs);
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
if (current_thread_info()->fpsaved[0] & flag) {
|
|
|
|
|
first = *(u64 *)&f->regs[freg];
|
|
|
|
|
second = *(u64 *)&f->regs[freg+2];
|
|
|
|
|
}
|
|
|
|
|
if (asi < 0x80) {
|
|
|
|
|
do_privact(regs);
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
switch (asi) {
|
|
|
|
|
case ASI_P:
|
|
|
|
|
case ASI_S: break;
|
|
|
|
|
case ASI_PL:
|
|
|
|
|
case ASI_SL:
|
|
|
|
|
{
|
|
|
|
|
/* Need to convert endians */
|
|
|
|
|
u64 tmp = __swab64p(&first);
|
|
|
|
|
|
|
|
|
|
first = __swab64p(&second);
|
|
|
|
|
second = tmp;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
default:
|
2006-02-10 04:20:34 +00:00
|
|
|
if (tlb_type == hypervisor)
|
|
|
|
|
sun4v_data_access_exception(regs, addr, 0);
|
|
|
|
|
else
|
|
|
|
|
spitfire_data_access_exception(regs, 0, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
if (put_user (first >> 32, (u32 __user *)addr) ||
|
|
|
|
|
__put_user ((u32)first, (u32 __user *)(addr + 4)) ||
|
|
|
|
|
__put_user (second >> 32, (u32 __user *)(addr + 8)) ||
|
|
|
|
|
__put_user ((u32)second, (u32 __user *)(addr + 12))) {
|
2006-02-10 04:20:34 +00:00
|
|
|
if (tlb_type == hypervisor)
|
|
|
|
|
sun4v_data_access_exception(regs, addr, 0);
|
|
|
|
|
else
|
|
|
|
|
spitfire_data_access_exception(regs, 0, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
/* LDF, LDDF, LDQF */
|
|
|
|
|
u32 data[4] __attribute__ ((aligned(8)));
|
|
|
|
|
int size, i;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
if (asi < 0x80) {
|
|
|
|
|
do_privact(regs);
|
|
|
|
|
return 1;
|
|
|
|
|
} else if (asi > ASI_SNFL) {
|
2006-02-10 04:20:34 +00:00
|
|
|
if (tlb_type == hypervisor)
|
|
|
|
|
sun4v_data_access_exception(regs, addr, 0);
|
|
|
|
|
else
|
|
|
|
|
spitfire_data_access_exception(regs, 0, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
switch (insn & 0x180000) {
|
|
|
|
|
case 0x000000: size = 1; break;
|
|
|
|
|
case 0x100000: size = 4; break;
|
|
|
|
|
default: size = 2; break;
|
|
|
|
|
}
|
sparc64: Don't restrict fp regs for no-fault loads
The function handle_ldf_stq() deals with no-fault ASI
loads and stores, but restricts fp registers to quad
word regs (ie, %f0, %f4 etc). This is valid for the
STQ case, but unnecessarily restricts loads, which
may be single precision, double, or quad. This results
in SIGFPE being raised for this instruction when the
source address is invalid:
ldda [%g1] ASI_PNF, %f2
but not for this one:
ldda [%g1] ASI_PNF, %f4
The validation check for quad register is moved to
within the STQ block so that loads are not affected
by the check.
An additional problem is that the calculation for freg
is incorrect when a single precision load is being
handled. This causes %f1 to be seen as %f32 etc,
and the incorrect register ends up being overwritten.
This code sequence demonstrates the problem:
ldd [%g1], %f32 ! g1 = valid address
lda [%i3] ASI_PNF, %f1 ! i3 = invalid address
std %f32, [%g1]
This is corrected by basing the freg calculation on
the load size.
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-10-31 04:36:23 +00:00
|
|
|
if (size == 1)
|
|
|
|
|
freg = (insn >> 25) & 0x1f;
|
|
|
|
|
else
|
|
|
|
|
freg = ((insn >> 25) & 0x1e) | ((insn >> 20) & 0x20);
|
|
|
|
|
flag = (freg < 32) ? FPRS_DL : FPRS_DU;
|
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
for (i = 0; i < size; i++)
|
|
|
|
|
data[i] = 0;
|
|
|
|
|
|
|
|
|
|
err = get_user (data[0], (u32 __user *) addr);
|
|
|
|
|
if (!err) {
|
|
|
|
|
for (i = 1; i < size; i++)
|
|
|
|
|
err |= __get_user (data[i], (u32 __user *)(addr + 4*i));
|
|
|
|
|
}
|
|
|
|
|
if (err && !(asi & 0x2 /* NF */)) {
|
2006-02-10 04:20:34 +00:00
|
|
|
if (tlb_type == hypervisor)
|
|
|
|
|
sun4v_data_access_exception(regs, addr, 0);
|
|
|
|
|
else
|
|
|
|
|
spitfire_data_access_exception(regs, 0, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
if (asi & 0x8) /* Little */ {
|
|
|
|
|
u64 tmp;
|
|
|
|
|
|
|
|
|
|
switch (size) {
|
|
|
|
|
case 1: data[0] = le32_to_cpup(data + 0); break;
|
|
|
|
|
default:*(u64 *)(data + 0) = le64_to_cpup((u64 *)(data + 0));
|
|
|
|
|
break;
|
|
|
|
|
case 4: tmp = le64_to_cpup((u64 *)(data + 0));
|
|
|
|
|
*(u64 *)(data + 0) = le64_to_cpup((u64 *)(data + 2));
|
|
|
|
|
*(u64 *)(data + 2) = tmp;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (!(current_thread_info()->fpsaved[0] & FPRS_FEF)) {
|
|
|
|
|
current_thread_info()->fpsaved[0] = FPRS_FEF;
|
|
|
|
|
current_thread_info()->gsr[0] = 0;
|
|
|
|
|
}
|
|
|
|
|
if (!(current_thread_info()->fpsaved[0] & flag)) {
|
|
|
|
|
if (freg < 32)
|
|
|
|
|
memset(f->regs, 0, 32*sizeof(u32));
|
|
|
|
|
else
|
|
|
|
|
memset(f->regs+32, 0, 32*sizeof(u32));
|
|
|
|
|
}
|
|
|
|
|
memcpy(f->regs + freg, data, size * 4);
|
|
|
|
|
current_thread_info()->fpsaved[0] |= flag;
|
|
|
|
|
}
|
|
|
|
|
advance(regs);
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void handle_ld_nf(u32 insn, struct pt_regs *regs)
|
|
|
|
|
{
|
|
|
|
|
int rd = ((insn >> 25) & 0x1f);
|
|
|
|
|
int from_kernel = (regs->tstate & TSTATE_PRIV) != 0;
|
|
|
|
|
unsigned long *reg;
|
|
|
|
|
|
2011-06-27 12:41:57 +00:00
|
|
|
perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0);
|
2009-12-11 09:07:53 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
maybe_flush_windows(0, 0, rd, from_kernel);
|
|
|
|
|
reg = fetch_reg_addr(rd, regs);
|
|
|
|
|
if (from_kernel || rd < 16) {
|
|
|
|
|
reg[0] = 0;
|
|
|
|
|
if ((insn & 0x780000) == 0x180000)
|
|
|
|
|
reg[1] = 0;
|
sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 22:18:37 +00:00
|
|
|
} else if (!test_thread_64bit_stack(regs->u_regs[UREG_FP])) {
|
2005-04-16 22:20:36 +00:00
|
|
|
put_user(0, (int __user *) reg);
|
|
|
|
|
if ((insn & 0x780000) == 0x180000)
|
|
|
|
|
put_user(0, ((int __user *) reg) + 1);
|
|
|
|
|
} else {
|
|
|
|
|
put_user(0, (unsigned long __user *) reg);
|
|
|
|
|
if ((insn & 0x780000) == 0x180000)
|
|
|
|
|
put_user(0, (unsigned long __user *) reg + 1);
|
|
|
|
|
}
|
|
|
|
|
advance(regs);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void handle_lddfmna(struct pt_regs *regs, unsigned long sfar, unsigned long sfsr)
|
|
|
|
|
{
|
2013-09-14 12:02:11 +00:00
|
|
|
enum ctx_state prev_state = exception_enter();
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long pc = regs->tpc;
|
|
|
|
|
unsigned long tstate = regs->tstate;
|
|
|
|
|
u32 insn;
|
|
|
|
|
u64 value;
|
2006-02-10 04:20:34 +00:00
|
|
|
u8 freg;
|
2005-04-16 22:20:36 +00:00
|
|
|
int flag;
|
|
|
|
|
struct fpustate *f = FPUSTATE;
|
|
|
|
|
|
|
|
|
|
if (tstate & TSTATE_PRIV)
|
|
|
|
|
die_if_kernel("lddfmna from kernel", regs);
|
2011-06-27 12:41:57 +00:00
|
|
|
perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, sfar);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (test_thread_flag(TIF_32BIT))
|
|
|
|
|
pc = (u32)pc;
|
|
|
|
|
if (get_user(insn, (u32 __user *) pc) != -EFAULT) {
|
2006-02-10 04:20:34 +00:00
|
|
|
int asi = decode_asi(insn, regs);
|
2009-01-09 00:52:36 +00:00
|
|
|
u32 first, second;
|
2009-01-08 01:15:57 +00:00
|
|
|
int err;
|
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if ((asi > ASI_SNFL) ||
|
|
|
|
|
(asi < ASI_P))
|
|
|
|
|
goto daex;
|
2009-01-09 00:52:36 +00:00
|
|
|
first = second = 0;
|
2009-01-08 01:15:57 +00:00
|
|
|
err = get_user(first, (u32 __user *)sfar);
|
|
|
|
|
if (!err)
|
|
|
|
|
err = get_user(second, (u32 __user *)(sfar + 4));
|
|
|
|
|
if (err) {
|
2009-01-09 00:52:36 +00:00
|
|
|
if (!(asi & 0x2))
|
2005-04-16 22:20:36 +00:00
|
|
|
goto daex;
|
2009-01-09 00:52:36 +00:00
|
|
|
first = second = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
save_and_clear_fpu();
|
|
|
|
|
freg = ((insn >> 25) & 0x1e) | ((insn >> 20) & 0x20);
|
|
|
|
|
value = (((u64)first) << 32) | second;
|
|
|
|
|
if (asi & 0x8) /* Little */
|
|
|
|
|
value = __swab64p(&value);
|
|
|
|
|
flag = (freg < 32) ? FPRS_DL : FPRS_DU;
|
|
|
|
|
if (!(current_thread_info()->fpsaved[0] & FPRS_FEF)) {
|
|
|
|
|
current_thread_info()->fpsaved[0] = FPRS_FEF;
|
|
|
|
|
current_thread_info()->gsr[0] = 0;
|
|
|
|
|
}
|
|
|
|
|
if (!(current_thread_info()->fpsaved[0] & flag)) {
|
|
|
|
|
if (freg < 32)
|
|
|
|
|
memset(f->regs, 0, 32*sizeof(u32));
|
|
|
|
|
else
|
|
|
|
|
memset(f->regs+32, 0, 32*sizeof(u32));
|
|
|
|
|
}
|
|
|
|
|
*(u64 *)(f->regs + freg) = value;
|
|
|
|
|
current_thread_info()->fpsaved[0] |= flag;
|
|
|
|
|
} else {
|
2006-02-10 04:20:34 +00:00
|
|
|
daex:
|
|
|
|
|
if (tlb_type == hypervisor)
|
|
|
|
|
sun4v_data_access_exception(regs, sfar, sfsr);
|
|
|
|
|
else
|
|
|
|
|
spitfire_data_access_exception(regs, sfsr, sfar);
|
2013-09-14 12:02:11 +00:00
|
|
|
goto out;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
advance(regs);
|
2013-09-14 12:02:11 +00:00
|
|
|
out:
|
|
|
|
|
exception_exit(prev_state);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void handle_stdfmna(struct pt_regs *regs, unsigned long sfar, unsigned long sfsr)
|
|
|
|
|
{
|
2013-09-14 12:02:11 +00:00
|
|
|
enum ctx_state prev_state = exception_enter();
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long pc = regs->tpc;
|
|
|
|
|
unsigned long tstate = regs->tstate;
|
|
|
|
|
u32 insn;
|
|
|
|
|
u64 value;
|
2006-02-10 04:20:34 +00:00
|
|
|
u8 freg;
|
2005-04-16 22:20:36 +00:00
|
|
|
int flag;
|
|
|
|
|
struct fpustate *f = FPUSTATE;
|
|
|
|
|
|
|
|
|
|
if (tstate & TSTATE_PRIV)
|
|
|
|
|
die_if_kernel("stdfmna from kernel", regs);
|
2011-06-27 12:41:57 +00:00
|
|
|
perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, 1, regs, sfar);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (test_thread_flag(TIF_32BIT))
|
|
|
|
|
pc = (u32)pc;
|
|
|
|
|
if (get_user(insn, (u32 __user *) pc) != -EFAULT) {
|
2006-02-10 04:20:34 +00:00
|
|
|
int asi = decode_asi(insn, regs);
|
2005-04-16 22:20:36 +00:00
|
|
|
freg = ((insn >> 25) & 0x1e) | ((insn >> 20) & 0x20);
|
|
|
|
|
value = 0;
|
|
|
|
|
flag = (freg < 32) ? FPRS_DL : FPRS_DU;
|
|
|
|
|
if ((asi > ASI_SNFL) ||
|
|
|
|
|
(asi < ASI_P))
|
|
|
|
|
goto daex;
|
|
|
|
|
save_and_clear_fpu();
|
|
|
|
|
if (current_thread_info()->fpsaved[0] & flag)
|
|
|
|
|
value = *(u64 *)&f->regs[freg];
|
|
|
|
|
switch (asi) {
|
|
|
|
|
case ASI_P:
|
|
|
|
|
case ASI_S: break;
|
|
|
|
|
case ASI_PL:
|
|
|
|
|
case ASI_SL:
|
|
|
|
|
value = __swab64p(&value); break;
|
|
|
|
|
default: goto daex;
|
|
|
|
|
}
|
|
|
|
|
if (put_user (value >> 32, (u32 __user *) sfar) ||
|
|
|
|
|
__put_user ((u32)value, (u32 __user *)(sfar + 4)))
|
|
|
|
|
goto daex;
|
|
|
|
|
} else {
|
2006-02-10 04:20:34 +00:00
|
|
|
daex:
|
|
|
|
|
if (tlb_type == hypervisor)
|
|
|
|
|
sun4v_data_access_exception(regs, sfar, sfsr);
|
|
|
|
|
else
|
|
|
|
|
spitfire_data_access_exception(regs, sfsr, sfar);
|
2013-09-14 12:02:11 +00:00
|
|
|
goto out;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
advance(regs);
|
2013-09-14 12:02:11 +00:00
|
|
|
out:
|
|
|
|
|
exception_exit(prev_state);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
