Hi Pete, Thanks for the info. A couple of very quick questions:
1) I presume this means that if there was some other user mode process running at the time of the interrupt, this other process will continue to run after the interrupt rather than waking up the blocked process? 2) If that other process blocks (before the internal scheduler timer ticks) than I presume the process that was triggered to wake up will wake up immediatley as well? If the answer is yes and yes, this information is very helpful for my continuing investigation with strace. I don't think I have no other background processes or operations that should be taking up 4 ms of CPU time. Even if those processes wake up they should finish their business and go to sleep rather quickly. Lastly, my assumption is that even if another process was running at the time of the interrupt (and does not block before the next timer interrupt), it will get preempted for the process that was originally woken up by the interrupt provided the process prorities are equal or greater for the woken up process. Best regards. p.s. I'm not sure about this reference thing but sorry for whatever inconvenience that is causing. I am using Yahoo Mail Beta which may be the cause for this consternation. It is a convenient way to anonymize though. ----- Original Message ---- From: Pete Zaitcev <[EMAIL PROTECTED]> To: Open Source <[EMAIL PROTECTED]> Cc: WolfgangMües <[EMAIL PROTECTED]>; linux-usb-devel@lists.sourceforge.net Sent: Saturday, October 14, 2006 12:58:33 PM Subject: Re: [linux-usb-devel] USB performance bug since kernel 2.6.13 (CRITICAL???) On Sat, 14 Oct 2006 11:07:37 -0700 (PDT), Open Source <[EMAIL PROTECTED]> wrote: > 4) The process "sleeps" to wait for the disk. > (No other processes are running so only the idle loop > is running at this point.) > 5) An interrupt triggers back from the disk. > 6) The process wakes up. > > Now if #6 didn't happen immediately, such user operations > would slow appreciably as CONFIG_HZ is reduced. > Maybe I misspoke by saying the "scheduler" runs for #6 > to wake up. I don't know the internals of the wait queue > infrastructure. But whatever it is, it has to wake up > immediately. It does wake up immediately if no other processes are running. Wolfgang simply forgot to mention that. The idle loop checks ->need_resched (or other equivalent mechanism in modern kernels). -- Pete P.S. I can understand that you want to hide, but at least get a decent mailer which keeps references. ------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ linux-usb-devel@lists.sourceforge.net To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
