2.5.47 scheduler problems?

Hi Mike, Rik, Everyone,

The O(1) schedule just isn't fair. It will run a subset of the runable processes excluding the rest. See my earlier emails for the details.

I had been working on a fix for this but got distracted by Posix timers. I still hope to get back to it.

My patch is here:
http://marc.theaimsgroup.com/?l=linux-kernel&m=103508412423719&w=2

It fixes fairness but breaks nice(2). Rik van Riel has a patch here which builds on my patch which fixes this:
http://marc.theaimsgroup.com/?l=linux-kernel&m=103651801424031&w=2

I just gave this a spin with. The patches still apply cleanly to linux-2.5.48 and it seems well behaved:-)

I found this problem with the LTP waitpid06 test. It actually produced a live-lock. See this mail:
http://marc.theaimsgroup.com/?l=linux-kernel&m=103133744217082&w=2

Jim Houston - Concurrent Computer Corp.

In a brief test, this seems to cure my problem.

(I haven't test this one yet)

It seems a little choppy still for a not swapping load, but greatly improved.

Thanks!

-Mike

(I put it into virgin 2.5.47 fwiw) I have some very odd behavior. I wanted to see how the kernel did at make -j30 bzImage on my test box to see what effect it has on throughput (box is 500 Mhz PIII + 128Mb ram), and get vmstat output like the attached. I should be roughly 30Mb into swap and paging heftily at this point.

-Mike

procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----

 b  w   swpd   free   buff  cache   si   so    bi    bo   in    cs us sy id wa
41  8    764  59940   1596  18484    0    0     4    27 1027   129 90 10  0  0
35 10    764  58932   1604  18616    0    0   120     0 1026   134 92  8  0  0
34  8    764  53540   1608  18804    0    0    29     0 1023   141 87 13  0  0
38  7    764  47780   1616  19300    0    0    10     0 1075   168 92  8  0  0
38 11    764  54476   1616  19684    0    0     8     0 1018   130 90 10  0  0
33 12    764  47596   1632  20048    0    0    14   474 1062   143 89 11  0  0
33  6    764  41564   1636  19412    0    0    88     0 1104   158 90 10  0  0
32  4    764  34940   1640  19728    0    0    93     0 1024   157 91  9  0  0
37  6    764  34652   1640  20012    0    0    38     0 1022   143 93  7  0  0
36  8    764  33196   1648  20108    0    0    18     0 1023   168 89 11  0  0
35  8    764  32668   1652  20528    0    0     8   110 1029   159 91  9  0  0
34  6    764  28364   1664  20852    0    0    84     1 1023   167 92  8  0  0
32  6    764  23140   1664  20608    0    0     4     0 1020   165 93  7  0  0
34  7    764  28372   1664  20424    0    0    12     0 1020   147 93  7  0  0
38  7    764  23476   1668  20480    0    0    45     0 1123   194 93  7  0  0
32  8    764  32988   1668  20952    0    0     4   132 1027   151 94  6  0  0
35  9    764  29260   1672  20332    0    0     4     1 1020   152 89 11  0  0
32  6    764  31100   1672  20688    0    0     0     0 1106   156 92  8  0  0
31  8    764  32076   1680  20264    0    0   108     0 1023   161 91  9  0  0
35  8    764  42332   1684  20836    0    0    17     0 1023   157 86 14  0  0
34 11    764  48916   1688  20752    0    0    52   234 1033   151 88 12  0  0
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
 b  w   swpd   free   buff  cache   si   so    bi    bo   in    cs us sy id wa
37  5    764  45796   1692  20956    0    0    12     1 1020   136 89 11  0  0
32  6    764  46820   1692  20616    0    0     0     0 1017   149 89 11  0  0
34  8    764  46820   1692  20652    0    0     0     0 1017   130 91  9  0  0
31  8    764  41316   1692  20496    0    0     8     0 1083   153 92  8  0  0

Never mind the vmstat output.. it seems you need both patches. With both in 2.5.48, the build progressed in a much more normal looking fashion. I'm not losing control of my box any more under load.

-Mike