Commit b1b5f65e53af770ede22c113e249de2f6fa53706
Committed by
Linus Torvalds
1 parent
191e56880a
Exists in
master
and in
4 other branches
calibrate: retry with wider bounds when converge seems to fail
Systems with unmaskable interrupts such as SMIs may massively underestimate loops_per_jiffy, and fail to converge anywhere near the real value. A case seen on x86_64 was an initial estimate of 256<<12, which converged to 511<<12 where the real value should have been over 630<<12. This admitedly requires bypassing the TSC calibration (lpj_fine), and a failure to settle in the direct calibration too, but is physically possible. This failure does not depend on my previous calibration optimisation, but by luck is easy to fix with the optimisation in place with a trivial retry loop. In the context of the optimised converging method, as we can no longer trust the starting estimate, enlarge the search bounds exponentially so that the number of retries is logarithmically bounded. [akpm@linux-foundation.org: mention x86_64 SMIs in comment] Signed-off-by: Phil Carmody <ext-phil.2.carmody@nokia.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Tested-by: Stephen Boyd <sboyd@codeaurora.org> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Showing 1 changed file with 18 additions and 4 deletions Side-by-side Diff
init/calibrate.c
... | ... | @@ -122,7 +122,7 @@ |
122 | 122 | static unsigned long __cpuinit calibrate_delay_converge(void) |
123 | 123 | { |
124 | 124 | /* First stage - slowly accelerate to find initial bounds */ |
125 | - unsigned long lpj, ticks, loopadd, chop_limit; | |
125 | + unsigned long lpj, lpj_base, ticks, loopadd, loopadd_base, chop_limit; | |
126 | 126 | int trials = 0, band = 0, trial_in_band = 0; |
127 | 127 | |
128 | 128 | lpj = (1<<12); |
129 | 129 | |
130 | 130 | |
... | ... | @@ -146,14 +146,18 @@ |
146 | 146 | * the largest likely undershoot. This defines our chop bounds. |
147 | 147 | */ |
148 | 148 | trials -= band; |
149 | - loopadd = lpj * band; | |
150 | - lpj *= trials; | |
151 | - chop_limit = lpj >> (LPS_PREC + 1); | |
149 | + loopadd_base = lpj * band; | |
150 | + lpj_base = lpj * trials; | |
152 | 151 | |
152 | +recalibrate: | |
153 | + lpj = lpj_base; | |
154 | + loopadd = loopadd_base; | |
155 | + | |
153 | 156 | /* |
154 | 157 | * Do a binary approximation to get lpj set to |
155 | 158 | * equal one clock (up to LPS_PREC bits) |
156 | 159 | */ |
160 | + chop_limit = lpj >> LPS_PREC; | |
157 | 161 | while (loopadd > chop_limit) { |
158 | 162 | lpj += loopadd; |
159 | 163 | ticks = jiffies; |
... | ... | @@ -164,6 +168,16 @@ |
164 | 168 | if (jiffies != ticks) /* longer than 1 tick */ |
165 | 169 | lpj -= loopadd; |
166 | 170 | loopadd >>= 1; |
171 | + } | |
172 | + /* | |
173 | + * If we incremented every single time possible, presume we've | |
174 | + * massively underestimated initially, and retry with a higher | |
175 | + * start, and larger range. (Only seen on x86_64, due to SMIs) | |
176 | + */ | |
177 | + if (lpj + loopadd * 2 == lpj_base + loopadd_base * 2) { | |
178 | + lpj_base = lpj; | |
179 | + loopadd_base <<= 2; | |
180 | + goto recalibrate; | |
167 | 181 | } |
168 | 182 | |
169 | 183 | return lpj; |