leandrof/linux
1
0
Fork 0
forked from Minki/linux
Code Issues Pull Requests Packages Projects Releases Wiki Activity
ae2163be10
linux/arch/powerpc/kernel/vdso32/gettimeofday.S

299 lines
6.9 KiB
ArmAsm
Raw Normal View History

[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
/*
* Userland implementation of gettimeofday() for 32 bits processes in a
* ppc64 kernel for use in the vDSO
*
* Copyright (C) 2004 Benjamin Herrenschmuidt (benh@kernel.crashing.org,
* IBM Corp.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*/
#include <asm/processor.h>
#include <asm/ppc_asm.h>
#include <asm/vdso.h>
#include <asm/asm-offsets.h>
#include <asm/unistd.h>
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
/* Offset for the low 32-bit part of a field of long type */
#ifdef CONFIG_PPC64
#define LOPART 4
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
#define TSPEC_TV_SEC TSPC64_TV_SEC+LOPART
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
#else
#define LOPART 0
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
#define TSPEC_TV_SEC TSPC32_TV_SEC
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
#endif
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
.text
/*
* Exact prototype of gettimeofday
*
* int __kernel_gettimeofday(struct timeval *tv, struct timezone *tz);
*
*/
V_FUNCTION_BEGIN(__kernel_gettimeofday)
.cfi_startproc
mflr r12
.cfi_register lr,r12
mr r10,r3 /* r10 saves tv */
mr r11,r4 /* r11 saves tz */
bl __get_datapage@local /* get data page */
mr r9, r3 /* datapage ptr in r9 */
[POWERPC] Fix VDSO gettimeofday() when called with NULL struct timeval Consider the prototype for gettimeofday(): int gettimofday(struct timeval *tv, struct timezone *tz); Although it is valid to call with /either/ tv or tz being NULL, and the C version of sys_gettimeofday() supports this, the current version of gettimeofday() in the VDSO will SEGV if called with a NULL tv. This adds a check for tv being NULL so that it doesn't SEGV. Signed-off-by: Tony Breeds <tony@bakeyournoodle.com> Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2007-06-25 23:50:32 +00:00
cmplwi r10,0 /* check if tv is NULL */
beq 3f
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
lis r7,1000000@ha /* load up USEC_PER_SEC */
addi r7,r7,1000000@l /* so we get microseconds in r4 */
bl __do_get_tspec@local /* get sec/usec from tb & kernel */
stw r3,TVAL32_TV_SEC(r10)
stw r4,TVAL32_TV_USEC(r10)
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
[POWERPC] Fix VDSO gettimeofday() when called with NULL struct timeval Consider the prototype for gettimeofday(): int gettimofday(struct timeval *tv, struct timezone *tz); Although it is valid to call with /either/ tv or tz being NULL, and the C version of sys_gettimeofday() supports this, the current version of gettimeofday() in the VDSO will SEGV if called with a NULL tv. This adds a check for tv being NULL so that it doesn't SEGV. Signed-off-by: Tony Breeds <tony@bakeyournoodle.com> Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2007-06-25 23:50:32 +00:00
3: cmplwi r11,0 /* check if tz is NULL */
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
beq 1f
lwz r4,CFG_TZ_MINUTEWEST(r9)/* fill tz */
lwz r5,CFG_TZ_DSTTIME(r9)
stw r4,TZONE_TZ_MINWEST(r11)
stw r5,TZONE_TZ_DSTTIME(r11)
1: mtlr r12
[PATCH] powerpc: Make the vDSO functions set error code (#2) The vDSO functions should have the same calling convention as a syscall. Unfortunately, they currently don't set the cr0.so bit which is used to indicate an error. This patch makes them clear this bit unconditionally since all functions currently succeed. The syscall fallback done by some of them will eventually override this if the syscall fails. This also changes the symbol version of all vdso exports to make sure glibc can differenciate between old and fixed calls for existing ones like __kernel_gettimeofday. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:54:32 +00:00
crclr cr0*4+so
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
li r3,0
blr
.cfi_endproc
V_FUNCTION_END(__kernel_gettimeofday)
/*
* Exact prototype of clock_gettime()
*
* int __kernel_clock_gettime(clockid_t clock_id, struct timespec *tp);
*
*/
V_FUNCTION_BEGIN(__kernel_clock_gettime)
.cfi_startproc
/* Check for supported clock IDs */
cmpli cr0,r3,CLOCK_REALTIME
cmpli cr1,r3,CLOCK_MONOTONIC
[PATCH] powerpc: vdso fixes (take #2) This fixes various errors in the new functions added in the vDSO's, I've now verified all functions on both 32 and 64 bits vDSOs. It also fix a sign extension bug getting the initial time of day at boot that could cause the monotonic clock value to be completely on bogus for 64 bits applications (with either the vDSO or the syscall) on powermacs. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-14 03:55:58 +00:00
cror cr0*4+eq,cr0*4+eq,cr1*4+eq
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
bne cr0,99f
mflr r12 /* r12 saves lr */
.cfi_register lr,r12
mr r11,r4 /* r11 saves tp */
bl __get_datapage@local /* get data page */
[PATCH] powerpc: vdso fixes (take #2) This fixes various errors in the new functions added in the vDSO's, I've now verified all functions on both 32 and 64 bits vDSOs. It also fix a sign extension bug getting the initial time of day at boot that could cause the monotonic clock value to be completely on bogus for 64 bits applications (with either the vDSO or the syscall) on powermacs. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-14 03:55:58 +00:00
mr r9,r3 /* datapage ptr in r9 */
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
lis r7,NSEC_PER_SEC@h /* want nanoseconds */
ori r7,r7,NSEC_PER_SEC@l
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
50: bl __do_get_tspec@local /* get sec/nsec from tb & kernel */
bne cr1,80f /* not monotonic -> all done */
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
/*
* CLOCK_MONOTONIC
*/
/* now we must fixup using wall to monotonic. We need to snapshot
* that value and do the counter trick again. Fortunately, we still
* have the counter value in r8 that was returned by __do_get_xsec.
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
* At this point, r3,r4 contain our sec/nsec values, r5 and r6
* can be used, r7 contains NSEC_PER_SEC.
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
*/
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
lwz r5,WTOM_CLOCK_SEC(r9)
lwz r6,WTOM_CLOCK_NSEC(r9)
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
/* We now have our offset in r5,r6. We create a fake dependency
* on that value and re-check the counter
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
*/
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
or r0,r6,r5
xor r0,r0,r0
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
add r9,r9,r0
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
lwz r0,(CFG_TB_UPDATE_COUNT+LOPART)(r9)
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
cmpl cr0,r8,r0 /* check if updated */
bne- 50b
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
/* Calculate and store result. Note that this mimics the C code,
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
* which may cause funny results if nsec goes negative... is that
* possible at all ?
*/
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
add r3,r3,r5
add r4,r4,r6
cmpw cr0,r4,r7
cmpwi cr1,r4,0
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
blt 1f
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
subf r4,r7,r4
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
addi r3,r3,1
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
1: bge cr1,80f
[PATCH] powerpc: vdso fixes (take #2) This fixes various errors in the new functions added in the vDSO's, I've now verified all functions on both 32 and 64 bits vDSOs. It also fix a sign extension bug getting the initial time of day at boot that could cause the monotonic clock value to be completely on bogus for 64 bits applications (with either the vDSO or the syscall) on powermacs. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-14 03:55:58 +00:00
addi r3,r3,-1
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
add r4,r4,r7
80: stw r3,TSPC32_TV_SEC(r11)
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
stw r4,TSPC32_TV_NSEC(r11)
mtlr r12
[PATCH] powerpc: Make the vDSO functions set error code (#2) The vDSO functions should have the same calling convention as a syscall. Unfortunately, they currently don't set the cr0.so bit which is used to indicate an error. This patch makes them clear this bit unconditionally since all functions currently succeed. The syscall fallback done by some of them will eventually override this if the syscall fails. This also changes the symbol version of all vdso exports to make sure glibc can differenciate between old and fixed calls for existing ones like __kernel_gettimeofday. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:54:32 +00:00
crclr cr0*4+so
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
li r3,0
blr
/*
* syscall fallback
*/
99:
li r0,__NR_clock_gettime
sc
blr
.cfi_endproc
V_FUNCTION_END(__kernel_clock_gettime)
/*
* Exact prototype of clock_getres()
*
* int __kernel_clock_getres(clockid_t clock_id, struct timespec *res);
*
*/
V_FUNCTION_BEGIN(__kernel_clock_getres)
.cfi_startproc
/* Check for supported clock IDs */
cmpwi cr0,r3,CLOCK_REALTIME
cmpwi cr1,r3,CLOCK_MONOTONIC
[PATCH] powerpc: vdso fixes (take #2) This fixes various errors in the new functions added in the vDSO's, I've now verified all functions on both 32 and 64 bits vDSOs. It also fix a sign extension bug getting the initial time of day at boot that could cause the monotonic clock value to be completely on bogus for 64 bits applications (with either the vDSO or the syscall) on powermacs. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-14 03:55:58 +00:00
cror cr0*4+eq,cr0*4+eq,cr1*4+eq
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
bne cr0,99f
li r3,0
cmpli cr0,r4,0
[PATCH] powerpc: Make the vDSO functions set error code (#2) The vDSO functions should have the same calling convention as a syscall. Unfortunately, they currently don't set the cr0.so bit which is used to indicate an error. This patch makes them clear this bit unconditionally since all functions currently succeed. The syscall fallback done by some of them will eventually override this if the syscall fails. This also changes the symbol version of all vdso exports to make sure glibc can differenciate between old and fixed calls for existing ones like __kernel_gettimeofday. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-16 02:54:32 +00:00
crclr cr0*4+so
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
beqlr
lis r5,CLOCK_REALTIME_RES@h
ori r5,r5,CLOCK_REALTIME_RES@l
stw r3,TSPC32_TV_SEC(r4)
stw r5,TSPC32_TV_NSEC(r4)
blr
/*
* syscall fallback
*/
99:
li r0,__NR_clock_getres
sc
blr
.cfi_endproc
V_FUNCTION_END(__kernel_clock_getres)
powerpc: Add VDSO version of time On 04/18/2013 07:38 PM, Anton Blanchard wrote: > Since you are only reading one long you shouldn't need to check the > update count and loop, you will always see a consistent value. The > system call version of time() just does an unprotected load for example. Fixed. > With the above change and with Michael's comments covered (decent > changelog entry and Signed-off-by): > > Acked-by: Anton Blanchard <anton@samba.org> Thanks for the review, below the updated patch: From: Adhemerval Zanella <azanella@linux.vnet.ibm.com> This patch implement the time syscall as vDSO. The performance speedups are: Baseline PPC32: 380 nsec Baseline PPC64: 350 nsec vdso PPC32: 20 nsec vsdo PPC64: 20 nsec Tested on 64 bit build with both 32 bit and 64 bit userland. Acked-by: Anton Blanchard <anton@samba.org> Signed-off-by: Adhemerval Zanella <azanella@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2013-04-22 09:29:33 +00:00
/*
* Exact prototype of time()
*
* time_t time(time *t);
*
*/
V_FUNCTION_BEGIN(__kernel_time)
.cfi_startproc
mflr r12
.cfi_register lr,r12
mr r11,r3 /* r11 holds t */
bl __get_datapage@local
mr r9, r3 /* datapage ptr in r9 */
lwz r3,STAMP_XTIME+TSPEC_TV_SEC(r9)
cmplwi r11,0 /* check if t is NULL */
beq 2f
stw r3,0(r11) /* store result at *t */
2: mtlr r12
crclr cr0*4+so
blr
.cfi_endproc
V_FUNCTION_END(__kernel_time)
[PATCH] powerpc: Merge vdso's and add vdso support to 32 bits kernel This patch moves the vdso's to arch/powerpc, adds support for the 32 bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds some new (still untested) routines to both vdso's: clock_gettime() with support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same clocks) and get_tbfreq() for glibc to retreive the timebase frequency. Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits returns a long long (r3, r4) not a long. This is such that if we ever add support for >4Ghz timebases on ppc32, the userland interface won't have to change. I have tested gettimeofday() using some glibc patches in both ppc32 and ppc64 kernels using 32 bits userland (I haven't had a chance to test a 64 bits userland yet, but the implementation didn't change and was tested earlier). I haven't tested yet the new functions. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2005-11-11 10:15:21 +00:00
/*
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
* This is the core of clock_gettime() and gettimeofday(),
* it returns the current time in r3 (seconds) and r4.
* On entry, r7 gives the resolution of r4, either USEC_PER_SEC
* or NSEC_PER_SEC, giving r4 in microseconds or nanoseconds.
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
* It expects the datapage ptr in r9 and doesn't clobber it.
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
* It clobbers r0, r5 and r6.
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
* On return, r8 contains the counter value that can be reused.
* This clobbers cr0 but not any other cr field.
*/
__do_get_tspec:
.cfi_startproc
/* Check for update count & load values. We use the low
* order 32 bits of the update count
*/
1: lwz r8,(CFG_TB_UPDATE_COUNT+LOPART)(r9)
andi. r0,r8,1 /* pending update ? loop */
bne- 1b
xor r0,r8,r8 /* create dependency */
add r9,r9,r0
/* Load orig stamp (offset to TB) */
lwz r5,CFG_TB_ORIG_STAMP(r9)
lwz r6,(CFG_TB_ORIG_STAMP+4)(r9)
/* Get a stable TB value */
powerpc/8xx: mfspr SPRN_TBRx in lieu of mftb/mftbu is not supported Commit beb2dc0a7a84be003ce54e98b95d65cc66e6e536 breaks the MPC8xx which seems to not support using mfspr SPRN_TBRx instead of mftb/mftbu despite what is written in the reference manual. This patch reverts to the use of mftb/mftbu when CONFIG_8xx is selected. Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr> Signed-off-by: Scott Wood <scottwood@freescale.com>
2013-11-22 16:57:31 +00:00
#ifdef CONFIG_8xx
2: mftbu r3
mftbl r4
mftbu r0
#else
powerpc: Convert some mftb/mftbu into mfspr Some CPUs (such as e500v1/v2) don't implement mftb and will take a trap. mfspr should work on everything that has a timebase, and is the preferred instruction according to ISA v2.06. Currently we get away with mftb on 85xx because the assembler converts it to mfspr due to -Wa,-me500. However, that flag has other effects that are undesireable for certain targets (e.g. lwsync is converted to sync), and is hostile to multiplatform kernels. Thus we would like to stop setting it for all e500-family builds. mftb/mftbu instances which are in 85xx code or common code are converted. Instances which will never run on 85xx are left alone. Signed-off-by: Scott Wood <scottwood@freescale.com>
2013-08-21 00:33:12 +00:00
2: mfspr r3, SPRN_TBRU
mfspr r4, SPRN_TBRL
mfspr r0, SPRN_TBRU
powerpc/8xx: mfspr SPRN_TBRx in lieu of mftb/mftbu is not supported Commit beb2dc0a7a84be003ce54e98b95d65cc66e6e536 breaks the MPC8xx which seems to not support using mfspr SPRN_TBRx instead of mftb/mftbu despite what is written in the reference manual. This patch reverts to the use of mftb/mftbu when CONFIG_8xx is selected. Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr> Signed-off-by: Scott Wood <scottwood@freescale.com>
2013-11-22 16:57:31 +00:00
#endif
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
cmplw cr0,r3,r0
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
bne- 2b
/* Subtract tb orig stamp and shift left 12 bits.
*/
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
subfc r4,r6,r4
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
subfe r0,r5,r3
slwi r0,r0,12
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
rlwimi. r0,r4,12,20,31
slwi r4,r4,12
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
/*
* Load scale factor & do multiplication.
* We only use the high 32 bits of the tb_to_xs value.
* Even with a 1GHz timebase clock, the high 32 bits of
* tb_to_xs will be at least 4 million, so the error from
* ignoring the low 32 bits will be no more than 0.25ppm.
* The error will just make the clock run very very slightly
* slow until the next time the kernel updates the VDSO data,
* at which point the clock will catch up to the kernel's value,
* so there is no long-term error accumulation.
*/
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
lwz r5,CFG_TB_TO_XS(r9) /* load values */
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
mulhwu r4,r4,r5
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
li r3,0
beq+ 4f /* skip high part computation if 0 */
mulhwu r3,r0,r5
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
mullw r5,r0,r5
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
addc r4,r4,r5
addze r3,r3
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
4:
/* At this point, we have seconds since the xtime stamp
* as a 32.32 fixed-point number in r3 and r4.
* Load & add the xtime stamp.
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
*/
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
lwz r5,STAMP_XTIME+TSPEC_TV_SEC(r9)
lwz r6,STAMP_SEC_FRAC(r9)
addc r4,r4,r6
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
adde r3,r3,r5
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
/* We create a fake dependency on the result in r3/r4
* and re-check the counter
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
*/
or r6,r4,r3
xor r0,r6,r6
add r9,r9,r0
lwz r0,(CFG_TB_UPDATE_COUNT+LOPART)(r9)
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
cmplw cr0,r8,r0 /* check if updated */
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
bne- 1b
powerpc: Rework VDSO gettimeofday to prevent time going backwards Currently it is possible for userspace to see the result of gettimeofday() going backwards by 1 microsecond, assuming that userspace is using the gettimeofday() in the VDSO. The VDSO gettimeofday() algorithm computes the time in "xsecs", which are units of 2^-20 seconds, or approximately 0.954 microseconds, using the algorithm now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec and then converts the time in xsecs to seconds and microseconds. The kernel updates the tb_orig_stamp and stamp_xsec values every tick in update_vsyscall(). If the length of the tick is not an integer number of xsecs, then some precision is lost in converting the current time to xsecs. For example, with CONFIG_HZ=1000, the tick is 1ms long, which is 1048.576 xsecs. That means that stamp_xsec will advance by either 1048 or 1049 on each tick. With the right conditions, it is possible for userspace to get (timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is slightly late in updating the vdso_datapage, and then for stamp_xsec to advance by 1048 when the kernel does update it, and for userspace to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to integer truncation. The result is that time appears to go backwards by 1 microsecond. To fix this we change the VDSO gettimeofday to use a new field in the VDSO datapage which stores the nanoseconds part of the time as a fractional number of seconds in a 0.32 binary fraction format. (Or put another way, as a 32-bit number in units of 0.23283 ns.) This is convenient because we can use the mulhwu instruction to convert it to either microseconds or nanoseconds. Since it turns out that computing the time of day using this new field is simpler than either using stamp_xsec (as gettimeofday does) or stamp_xtime.tv_nsec (as clock_gettime does), this converts both gettimeofday and clock_gettime to use the new field. The existing __do_get_tspec function is converted to use the new field and take a parameter in r7 that indicates the desired resolution, 1,000,000 for microseconds or 1,000,000,000 for nanoseconds. The __do_get_xsec function is then unused and is deleted. The new algorithm is now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs + (stamp_xtime_seconds << 32) + stamp_sec_fraction with 'now' in units of 2^-32 seconds. That is then converted to seconds and either microseconds or nanoseconds with seconds = now >> 32 partseconds = ((now & 0xffffffff) * resolution) >> 32 The 32-bit VDSO code also makes a further simplification: it ignores the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary fraction. Doing so gets rid of 4 multiply instructions. Assuming a timebase frequency of 1GHz or less and an update interval of no more than 10ms, the upper 32 bits of tb_to_xs will be at least 4503599, so the error from ignoring the low 32 bits will be at most 2.2ns, which is more than an order of magnitude less than the time taken to do gettimeofday or clock_gettime on our fastest processors, so there is no possibility of seeing inconsistent values due to this. This also moves update_gtod() down next to its only caller, and makes update_vsyscall use the time passed in via the wall_time argument rather than accessing xtime directly. At present, wall_time always points to xtime, but that could change in future. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-06-20 19:03:08 +00:00
mulhwu r4,r4,r7 /* convert to micro or nanoseconds */
powerpc: Improve resolution of VDSO clock_gettime Currently the clock_gettime implementation in the VDSO produces a result with microsecond resolution for the cases that are handled without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The nanoseconds field of the result is obtained by computing a microseconds value and multiplying by 1000. This changes the code in the VDSO to do the computation for clock_gettime with nanosecond resolution. That means that the resolution of the result will ultimately depend on the timebase frequency. Because the timestamp in the VDSO datapage (stamp_xsec, the real time corresponding to the timebase count in tb_orig_stamp) is in units of 2^-20 seconds, it doesn't have sufficient resolution for computing a result with nanosecond resolution. Therefore this adds a copy of xtime to the VDSO datapage and updates it in update_gtod() along with the other time-related fields. Signed-off-by: Paul Mackerras <paulus@samba.org>
2008-10-27 23:56:03 +00:00
blr
.cfi_endproc
Reference in New Issue Copy Permalink