diff options
author | Stephane Eranian <eranian@hpl.hp.com> | 2005-04-11 16:45:00 -0400 |
---|---|---|
committer | Tony Luck <tony.luck@intel.com> | 2005-05-03 18:44:48 -0400 |
commit | 8df5a500a3e97f7811cdce0f553ca1917ccd4220 (patch) | |
tree | cac08cc58a5e9fb846a9946a52026fcbe7718634 /arch | |
parent | 012914dad25bd5cacf88af4429eecda62a06020d (diff) |
[IA64] perfmon & PAL_HALT again
The pmu_active test is based on the values of PSR.up. THIS IS THE PROBLEM as
it does not take into account the lazy restore logic which is as follow (simplified):
context switch out:
save PMDs
clear psr.up
release ownership
context switch in:
if (ctx->last_cpu == smp_processor_id() && ctx->cpu_activation == cpu_activation) {
set psr.up
return
}
restore PMD
restore PMC
ctx->last_cpu = smp_processor_id();
ctx->activation = ++cpu_activation;
set psr.up
The key here is that on context switch out, we clear psr.up and on context switch in
we check if nobody else used the PMU on that processor since last time we came. In
that case, we assume the PMD/PMC are ours and we simply reactivate.
The Caliper problem is that between the moment we context switch out and the moment we
come back, nobody effectively used the PMU BUT the processor went idle. Normally this
would have no incidence but PAL_HALT does alter the PMU registers. In default_idle(),
the test on psr.up is not strong enough to cover this case and we go into PAL which
trashed the PMU resgisters. When we come back we falsely assume that this is our state
yet it is corrupted. Very nasty indeed.
To avoid the problem it is necessary to forbid going to PAL_HALT as soon as perfmon
installs some valid state in the PMU registers. This happens with an application
attaches a context to a thread or CPU. It is not enough to check the psr/dcr bits.
Hence I propose the attached patch. It adds a callback in process.c to modify the
condition to enter PAL on idle. Basically, now it is conditional to pal_halt=1 AND
perfmon saying it is okay.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Diffstat (limited to 'arch')
-rw-r--r-- | arch/ia64/kernel/perfmon.c | 13 | ||||
-rw-r--r-- | arch/ia64/kernel/process.c | 14 |
2 files changed, 24 insertions, 3 deletions
diff --git a/arch/ia64/kernel/perfmon.c b/arch/ia64/kernel/perfmon.c index 376fcbc3f8da..fd4f3be6e856 100644 --- a/arch/ia64/kernel/perfmon.c +++ b/arch/ia64/kernel/perfmon.c | |||
@@ -1265,6 +1265,8 @@ out: | |||
1265 | } | 1265 | } |
1266 | EXPORT_SYMBOL(pfm_unregister_buffer_fmt); | 1266 | EXPORT_SYMBOL(pfm_unregister_buffer_fmt); |
1267 | 1267 | ||
1268 | extern void update_pal_halt_status(int); | ||
1269 | |||
1268 | static int | 1270 | static int |
1269 | pfm_reserve_session(struct task_struct *task, int is_syswide, unsigned int cpu) | 1271 | pfm_reserve_session(struct task_struct *task, int is_syswide, unsigned int cpu) |
1270 | { | 1272 | { |
@@ -1311,6 +1313,11 @@ pfm_reserve_session(struct task_struct *task, int is_syswide, unsigned int cpu) | |||
1311 | is_syswide, | 1313 | is_syswide, |
1312 | cpu)); | 1314 | cpu)); |
1313 | 1315 | ||
1316 | /* | ||
1317 | * disable default_idle() to go to PAL_HALT | ||
1318 | */ | ||
1319 | update_pal_halt_status(0); | ||
1320 | |||
1314 | UNLOCK_PFS(flags); | 1321 | UNLOCK_PFS(flags); |
1315 | 1322 | ||
1316 | return 0; | 1323 | return 0; |
@@ -1366,6 +1373,12 @@ pfm_unreserve_session(pfm_context_t *ctx, int is_syswide, unsigned int cpu) | |||
1366 | is_syswide, | 1373 | is_syswide, |
1367 | cpu)); | 1374 | cpu)); |
1368 | 1375 | ||
1376 | /* | ||
1377 | * if possible, enable default_idle() to go into PAL_HALT | ||
1378 | */ | ||
1379 | if (pfm_sessions.pfs_task_sessions == 0 && pfm_sessions.pfs_sys_sessions == 0) | ||
1380 | update_pal_halt_status(1); | ||
1381 | |||
1369 | UNLOCK_PFS(flags); | 1382 | UNLOCK_PFS(flags); |
1370 | 1383 | ||
1371 | return 0; | 1384 | return 0; |
diff --git a/arch/ia64/kernel/process.c b/arch/ia64/kernel/process.c index c0140f4235e4..474d75f9de8a 100644 --- a/arch/ia64/kernel/process.c +++ b/arch/ia64/kernel/process.c | |||
@@ -173,7 +173,9 @@ do_notify_resume_user (sigset_t *oldset, struct sigscratch *scr, long in_syscall | |||
173 | ia64_do_signal(oldset, scr, in_syscall); | 173 | ia64_do_signal(oldset, scr, in_syscall); |
174 | } | 174 | } |
175 | 175 | ||
176 | static int pal_halt = 1; | 176 | static int pal_halt = 1; |
177 | static int can_do_pal_halt = 1; | ||
178 | |||
177 | static int __init nohalt_setup(char * str) | 179 | static int __init nohalt_setup(char * str) |
178 | { | 180 | { |
179 | pal_halt = 0; | 181 | pal_halt = 0; |
@@ -181,16 +183,22 @@ static int __init nohalt_setup(char * str) | |||
181 | } | 183 | } |
182 | __setup("nohalt", nohalt_setup); | 184 | __setup("nohalt", nohalt_setup); |
183 | 185 | ||
186 | int | ||
187 | update_pal_halt_status(int status) | ||
188 | { | ||
189 | can_do_pal_halt = pal_halt && status; | ||
190 | } | ||
191 | |||
184 | /* | 192 | /* |
185 | * We use this if we don't have any better idle routine.. | 193 | * We use this if we don't have any better idle routine.. |
186 | */ | 194 | */ |
187 | void | 195 | void |
188 | default_idle (void) | 196 | default_idle (void) |
189 | { | 197 | { |
190 | unsigned long pmu_active = ia64_getreg(_IA64_REG_PSR) & (IA64_PSR_PP | IA64_PSR_UP); | 198 | int can_do_pal; |
191 | 199 | ||
192 | while (!need_resched()) | 200 | while (!need_resched()) |
193 | if (pal_halt && !pmu_active) | 201 | if (can_do_pal_halt) |
194 | safe_halt(); | 202 | safe_halt(); |
195 | else | 203 | else |
196 | cpu_relax(); | 204 | cpu_relax(); |