search for: smap

...heads like checks, spec barriers or even hardware feature >>>> toggling. >>> Will review, thanks! >>> One questions that comes to mind is whether it's all about bypassing >>> stac/clac. Could you please include a performance comparison with >>> nosmap? >>> >> On machine without SMAP (Sandy Bridge): >> >> Before: 4.8Mpps >> >> After: 5.2Mpps > OK so would you say it's really unsafe versus safe accesses? > Or would you say it's just a better written code? It's the effect of removing specul...

...heads like checks, spec barriers or even hardware feature >>>> toggling. >>> Will review, thanks! >>> One questions that comes to mind is whether it's all about bypassing >>> stac/clac. Could you please include a performance comparison with >>> nosmap? >>> >> On machine without SMAP (Sandy Bridge): >> >> Before: 4.8Mpps >> >> After: 5.2Mpps > OK so would you say it's really unsafe versus safe accesses? > Or would you say it's just a better written code? It's the effect of removing specul...

...ds since they had too much >> overheads like checks, spec barriers or even hardware feature >> toggling. > Will review, thanks! > One questions that comes to mind is whether it's all about bypassing > stac/clac. Could you please include a performance comparison with > nosmap? > On machine without SMAP (Sandy Bridge): Before: 4.8Mpps After: 5.2Mpps On machine with SMAP (Broadwell): Before: 5.0Mpps After: 6.1Mpps No smap: 7.5Mpps Thanks

...ds since they had too much >> overheads like checks, spec barriers or even hardware feature >> toggling. > Will review, thanks! > One questions that comes to mind is whether it's all about bypassing > stac/clac. Could you please include a performance comparison with > nosmap? > On machine without SMAP (Sandy Bridge): Before: 4.8Mpps After: 5.2Mpps On machine with SMAP (Broadwell): Before: 5.0Mpps After: 6.1Mpps No smap: 7.5Mpps Thanks

This appears to be the case in the recent messages in Debian bug #660425 [0] (thread starts at [1] and continues at [2]). SMAP uses EFLAGS.AC in kernel mode to indicate whether accesses to usermode pages should fault or not, which gets confused because under 64-bit Xen kernel == ring3. The obvious fix would be to clear SMAP from the cpuid features (probably in kernel as well as hypervosr) but might we want to do something...

...> > > > > > virtqueue metdata tends to be very expensive for dataplane > > > > > > > > > implementation like vhost since it involves lots of software check, > > > > > > > > > speculation barrier, hardware feature toggling (e.g SMAP). The > > > > > > > > > extra cost will be more obvious when transferring small packets. > > > > > > > > > > > > > > > > > > This patch tries to eliminate those overhead by pin vq metadata pages > > > > &g...

...> > > > > > virtqueue metdata tends to be very expensive for dataplane > > > > > > > > > implementation like vhost since it involves lots of software check, > > > > > > > > > speculation barrier, hardware feature toggling (e.g SMAP). The > > > > > > > > > extra cost will be more obvious when transferring small packets. > > > > > > > > > > > > > > > > > > This patch tries to eliminate those overhead by pin vq metadata pages > > > > &g...

...used to access > > > > > > > virtqueue metdata tends to be very expensive for dataplane > > > > > > > implementation like vhost since it involves lots of software check, > > > > > > > speculation barrier, hardware feature toggling (e.g SMAP). The > > > > > > > extra cost will be more obvious when transferring small packets. > > > > > > > > > > > > > > This patch tries to eliminate those overhead by pin vq metadata pages > > > > > > > and access them t...

...used to access > > > > > > > virtqueue metdata tends to be very expensive for dataplane > > > > > > > implementation like vhost since it involves lots of software check, > > > > > > > speculation barrier, hardware feature toggling (e.g SMAP). The > > > > > > > extra cost will be more obvious when transferring small packets. > > > > > > > > > > > > > > This patch tries to eliminate those overhead by pin vq metadata pages > > > > > > > and access them t...

...> > virtqueue metdata tends to be very expensive for dataplane > > > > > > > > > > > implementation like vhost since it involves lots of software check, > > > > > > > > > > > speculation barrier, hardware feature toggling (e.g SMAP). The > > > > > > > > > > > extra cost will be more obvious when transferring small packets. > > > > > > > > > > > > > > > > > > > > > > This patch tries to eliminate those overhead by pin vq metadat...

...even hardware feature > > > > > toggling. > > > > Will review, thanks! > > > > One questions that comes to mind is whether it's all about bypassing > > > > stac/clac. Could you please include a performance comparison with > > > > nosmap? > > > > > > > On machine without SMAP (Sandy Bridge): > > > > > > Before: 4.8Mpps > > > > > > After: 5.2Mpps > > OK so would you say it's really unsafe versus safe accesses? > > Or would you say it's just a better writ...

...[DM5020, MCK-RIPE, DM3-RIPN] ------------------------------------------------------------------------ *** Dmitry Morozovsky --- D.Marck --- Wild Woozle --- marck@rinet.ru *** ------------------------------------------------------------------------ -------------- next part -------------- ok boot -sv SMAP type=01 base=00000000 00000000 len=00000000 0009fc00 SMAP type=02 base=00000000 0009fc00 len=00000000 00000400 SMAP type=02 base=00000000 000f0000 len=00000000 00010000 SMAP type=01 base=00000000 00100000 len=00000000 1fdfa000 SMAP type=03 base=00000000 1fefa000 len=00000000 00003000 SMAP type=04 b...

...t;>>>>>>>>> virtqueue metdata tends to be very expensive for dataplane >>>>>>>>>> implementation like vhost since it involves lots of software check, >>>>>>>>>> speculation barrier, hardware feature toggling (e.g SMAP). The >>>>>>>>>> extra cost will be more obvious when transferring small packets. >>>>>>>>>> >>>>>>>>>> This patch tries to eliminate those overhead by pin vq metadata pages >>>>>>>>&gt...

...at was used to access >>>>>>>> virtqueue metdata tends to be very expensive for dataplane >>>>>>>> implementation like vhost since it involves lots of software check, >>>>>>>> speculation barrier, hardware feature toggling (e.g SMAP). The >>>>>>>> extra cost will be more obvious when transferring small packets. >>>>>>>> >>>>>>>> This patch tries to eliminate those overhead by pin vq metadata pages >>>>>>>> and access them through vm...

...eing it reliably on a RELENG_6 checkout from 5th of May. It affects multiple (but identical) systems, here's an verbose boot leading to the panic. Please note that 6.2 was running fine on these machines, they also boot "normally" if I disable ACPI (but this is not really an option). SMAP type=01 base=0000000000000000 len=000000000009d800 SMAP type=02 base=000000000009d800 len=0000000000002800 SMAP type=02 base=00000000000ce000 len=0000000000002000 SMAP type=02 base=00000000000e4000 len=000000000001c000 SMAP type=01 base=0000000000100000 len=00000000cfe60000 SMAP type=03 base=000000...

Hi all, I have again noticed that the wiki does not really show up in search results and wonder if it has any impact that robots.txt on wiki.centos.org is empty. Perhaps it should at least contain User-agent: * ? Best Regards Marcus

...e copy_user() friends that was used to access > > > > > virtqueue metdata tends to be very expensive for dataplane > > > > > implementation like vhost since it involves lots of software check, > > > > > speculation barrier, hardware feature toggling (e.g SMAP). The > > > > > extra cost will be more obvious when transferring small packets. > > > > > > > > > > This patch tries to eliminate those overhead by pin vq metadata pages > > > > > and access them through vmap(). During SET_VRING_ADDR, we...

...e copy_user() friends that was used to access > > > > > virtqueue metdata tends to be very expensive for dataplane > > > > > implementation like vhost since it involves lots of software check, > > > > > speculation barrier, hardware feature toggling (e.g SMAP). The > > > > > extra cost will be more obvious when transferring small packets. > > > > > > > > > > This patch tries to eliminate those overhead by pin vq metadata pages > > > > > and access them through vmap(). During SET_VRING_ADDR, we...

...0, Sven Schwedas wrote: > On 2017-05-05 10:09, Volker Lendecke wrote: > > On Fri, May 05, 2017 at 09:42:47AM +0200, Sven Schwedas via samba wrote: > >>> root 9988 0.8 59.4 1571936 606488 ? S Apr26 114:41 /usr/sbin/samba > > > > Can you post /proc/9988/smaps somewhere? > > Sure, http://up.tao.at/-0242609d03e69298bb27155078663d51/smaps.9988.txt Ok, it's actually the heap. I was hoping that it's a mmap'ed ldb. So we do have something fishy going on that. Sorry, I con't have an easy recipe. Volker -- SerNet GmbH, Bahnhofsallee...

...t; > It was noticed that the copy_user() friends that was used to access > > > virtqueue metdata tends to be very expensive for dataplane > > > implementation like vhost since it involves lots of software check, > > > speculation barrier, hardware feature toggling (e.g SMAP). The > > > extra cost will be more obvious when transferring small packets. > > > > > > This patch tries to eliminate those overhead by pin vq metadata pages > > > and access them through vmap(). During SET_VRING_ADDR, we will setup > > > those mapping...