search for: emption

...r), it's slightly cleaner if we instead declare that we are already blocked. I tested with: $ nbdkit -U - null 16M --run 'examples/threaded-reads-and-writes $unixsocket' There was no real noticeable difference in timing, but I did observe that pre-patch, the run encountered 168825 pre-emptions and 136976 send() EAGAIN failures (81%), while post-patch the run encountered 166066 pre-emptions and 552 EAGAIN failures (0.3%). --- generator/states-issue-command.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) diff --git a/generator/states-issue-command.c b/generator/state...

Besides playing with Asterisk, i'm also using Linux for all kinds of multimedia things, especially recording music, mixing, etc. In order to use Linux as a digital audio workstation, there are a few things that one must do: use low-latency kernels, use pre-emption, use apps that run with real-time privileges, etc. For example, among audio Linux users, the CK (Con Kolivas) and LCK (Locosoft CK) patches are popular: http://members.optusnet.com.au/ckolivas/kernel/ http://www.plumlocosoft.com/kernel/ These patches provide O(1) scheduler, pre-emption, low late...

...> instead declare that we are already blocked. > > I tested with: > $ nbdkit -U - null 16M --run 'examples/threaded-reads-and-writes $unixsocket' > > There was no real noticeable difference in timing, but I did observe > that pre-patch, the run encountered 168825 pre-emptions and 136976 > send() EAGAIN failures (81%), while post-patch the run encountered > 166066 pre-emptions and 552 EAGAIN failures (0.3%). > --- > generator/states-issue-command.c | 11 ++++++++--- > 1 file changed, 8 insertions(+), 3 deletions(-) > > diff --git a/generator/state...

When processing a server reply within the REPLY subgroup, we will often hit a situation where recv() requires us to block until the next NotifyRead. But since NotifyRead is the only permitted external action while in this group, we are effectively blocking CmdIssue and NotifyWrite events from happening until the server finishes the in-progress reply, even though those events have no strict

...if this matters. But we have also chinese students and > researchers all abroad. Nearly an intractable problem, especially since your users are embedded in a notoriously infested network (as someone quipped, "like picking marshmallows out from a pile of sh*t"). Some ideas: - pre-emption (using third party RBLs that targets BFD) - immediate blacklisting of known bad users/passwords (e.g. "admin", "support", extinct users, etc.) - persistent tracking storage: tracking in SQL, or or large LRU list that can reach far enough back in time. (I think Aki...

...ent of WINE. They place certain system services in the kernel and have compile kernel-module (.ko) to service them. The kernel patches are very small so this is not "bloat" as someone has stated. My main question is how such patches effect OTHER common patches, especially real-time pre-emption. If this be safe, I have no objection to trying it. The "unified kernel" project patches WINE to use its kernel module, otherwise, UTILIZES WINE for its API layer. Since WINE is changed quite often, continually patching and (re-) compiling is a lot of effort and maybe not really wor...

We have setup asterisk to handle our calls before between telco and an Avaya definity. The PRI keeps locking up every so often. In addition I keep getting this error when trying to call the avaya: -- Channel 0/2, span 1 got hangup request, cause 102 -- Hungup 'Zap/2-1' When that error happens I get a fast busy (congestion) tone. Any one can point me in the right direction? TIA

...or use in this call. 7 07 Call awarded and being delivered in an Established channel.Indicates that the user has been awarded the incoming call, and that the incoming call is being connected to a channel already established to that user for similar calls (e.g. packet-mode virtual calls). 8 08 Pre-Emption. Indicates that the call is being pre-empted. 0 09 Pre-Emption Circuit reserved for reuse. Indicates that the call is being cleared because one of the users involved in the call has requested that the call be cleared. 16 10 Normal Call Clearing. This is one of the most common cause codes and is r...

Hello all, As some of you may already know from comments in the GitHub pull requests and other media, due to pre-emption by real life our long-term leader and main maintainer, Arnaud Quette, had to step down from this role and from intensive involvement in the NUT project (we hope he would still be able to help on occasion, time permitting). He asked the members of the core team to take over, and it seems I am the on...

Hello all, As some of you may already know from comments in the GitHub pull requests and other media, due to pre-emption by real life our long-term leader and main maintainer, Arnaud Quette, had to step down from this role and from intensive involvement in the NUT project (we hope he would still be able to help on occasion, time permitting). He asked the members of the core team to take over, and it seems I am the on...

...; busyloop delay burning CPU cycles for nothing. msleep() is not, > > it will reschedule until the delay has passed. The change > > raises two questions: > > > > - Is the bios code ever called in a situation, where scheduling > > is not allowed (interrupt context, pre-emption disabled, inside > > spinlocks, etc.)? > > not as far as I know, though I'm uncertain of what the conditions > are when resuming from suspend > > > - Is the delay accuracy on the msleep path critical? > > no (but on udelay it possibly is) > > btw, does mi...

Hi, questions will follow. --- diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.c b/drivers/gpu/drm/nouveau/nouveau_bios.c index 99f7bd4..13b3fb1 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bios.c +++ b/drivers/gpu/drm/nouveau/nouveau_bios.c @@ -40,8 +40,6 @@ #define BIOSLOG(sip, fmt, arg...) NV_DEBUG(sip->dev, fmt, ##arg) #define LOG_OLD_VALUE(x) //x -#define BIOS_USLEEP(n)

...age fault handler, it can = be pre-empted. Both VMI and Xen use per-cpu variables to track lazy = mode state, as all previous calls to set, disable, or flush lazy mode = happened from a non-preemptable state. I have no idea how to convincingly produce the problem, as generating a = kernel pre-emption at the required point is, um, difficult, but it is = most certainly a real possibility, and potentially more likely than the = bug I fixed originally. Rusty, you may have to modify lguest code if you use lazy mode and rely = on per-cpu variables during the callout for paravirt_ops.set_lazy_mode...

...age fault handler, it can = be pre-empted. Both VMI and Xen use per-cpu variables to track lazy = mode state, as all previous calls to set, disable, or flush lazy mode = happened from a non-preemptable state. I have no idea how to convincingly produce the problem, as generating a = kernel pre-emption at the required point is, um, difficult, but it is = most certainly a real possibility, and potentially more likely than the = bug I fixed originally. Rusty, you may have to modify lguest code if you use lazy mode and rely = on per-cpu variables during the callout for paravirt_ops.set_lazy_mode...

What are methods to debug domU when it hangs? Using xen 3.1 compiled from sources I could not manage to launch no domU. Fot instance, I run something like this: =====8<===================== disk = [ ''file:/oradata-act/sles.disk,hda1,w'', '',hdc:cdrom,r'' ] kernel = "/boot/vmlinuz-2.6.18-xen" ramdisk = "/boot/initrd-2.6.18-xen" cpus =

...g(struct task_ return retval; } -static unsigned long convert_eip_to_linear(struct task_struct *child, struct pt_regs *regs) +/* + * Get the GDT/LDT descriptor base. When you look for races in this code + * remember that LDT and other horrors are only used in user space. Must + * disable pre-emption to reading the GDT, and must take the LDT semaphore + * for LDT segments. The fast path handles standard kernel and user CS + * as well as V8086 mode. + */ +unsigned long convert_eip_to_linear_slow(unsigned long eip, unsigned long seg, + mm_context_t *context, unsigned long *eip_limit) { - unsig...

...g(struct task_ return retval; } -static unsigned long convert_eip_to_linear(struct task_struct *child, struct pt_regs *regs) +/* + * Get the GDT/LDT descriptor base. When you look for races in this code + * remember that LDT and other horrors are only used in user space. Must + * disable pre-emption to reading the GDT, and must take the LDT semaphore + * for LDT segments. The fast path handles standard kernel and user CS + * as well as V8086 mode. + */ +unsigned long convert_eip_to_linear_slow(unsigned long eip, unsigned long seg, + mm_context_t *context, unsigned long *eip_limit) { - unsig...

...only valid - // registers are the 4 used for parameters. We don't currently do this - // case. - if (Subtarget->isThumb1Only()) - return false; - // Externally-defined functions with weak linkage should not be // tail-called on ARM when the OS does not support dynamic // pre-emption of symbols, as the AAELF spec requires normal calls @@ -2365,7 +2376,7 @@ if (!CI->isTailCall() || getTargetMachine().Options.DisableTailCalls) return false; - return !Subtarget->isThumb1Only(); + return true; } // ConstantPool, JumpTable, GlobalAddress, and ExternalSymbol are...

...tions with the guest OSes internal scheduler. Currently, no attempt is made to gang schedule VCPUs belonging to the same guest. For most workloads this seems to work OK. In future we may have to investigate schemes that switch to gang scheduling under certain circumstances, or otherwise use bad pre-emption avoidance (e.g. don''t pre-empt while kernel locks held) or bad pre-emption mitigation strategies (e.g. directed yield). It may be possible to dynamically spot groups of VCPUs that are actively communicating and gang schedule them. We will need to extend the scheduler to better understand C...