Bug#556030: Root cause: deadlock in a driver

2011-08-26 Thread François-Frédéric Ozog
We extensively researched the problem. The TLB flush softlockup is only a CONSEQUENCE of a deadlock. Background: The TLB flush is issued by a CPU to a number of other CPUs using inter-processor interupts to progagate paging changes. Then the issuing CPU loops until all processor acknowledg

Bug#556030: native_flush_tlb_others BUG on 2.6.32.7

2010-05-03 Thread François-Frédéric Ozog
I posted the bug as per Moritz Muehlenhoff and Maximilian Attems requests on the same thread to report on such version. On Mon, 2010-05-03 at 12:42 +0100, Ben Hutchings wrote: > On Mon, 2010-05-03 at 10:45 +0200, François-Frédéric Ozog wrote: > > We started to get those softockups sinc

Bug#556030: native_flush_tlb_others BUG on 2.6.32.7

2010-05-03 Thread François-Frédéric Ozog
We started to get those softockups since we used 2.6.32.7. We had no such issues on 2.6.29.5 on the very same hardware. [ 6698.002377] BUG: soft lockup - CPU#0 stuck for 31s! [lighttpd:3894] [ 6698.009802] Modules linked in: plm cxgb3 e1000e pfm vnetdev binfmt_misc cpufreq_ondemand freq_table ipm