Re: [Xenomai] [RFC] RTnet fixes and other updates
On 01/24/2018 09:44 AM, Leopold Palomo-Avellaneda wrote: > Philippe, > > > On 23/01/18 18:21, Philippe Gerum wrote: > >> For review by interested parties, proposed for inclusion in 3.0.6. > > 3.0.7 ? > Yep. Time is not moving backward - yet. -- Philippe. ___ Xenomai mailing list Xenomai@xenomai.org https://xenomai.org/mailman/listinfo/xenomai
Re: [Xenomai] ipipe: kernel updates in the light of recent fixes (meltdown...)
Hi Jan, On 01/23/2018 06:52 PM, Jan Kiszka wrote: > Hi Philippe, > > as the number of question increase here, but I bet elsewhere as well, > and I'm planning to work on some parts soonish, I'd like to coordinate: > > Where are we standing with 4.14 right now? There is nothing for x86 yet, > right? > Correct. > What are the update plans for 4.9? We currently have one x86 user on > that kernel, but I suppose we could also move on to 4.14 once it's > ready. For me the question is right now where to invest x86-wise, 4.9 > update or 4.14 completion + update? > Regarding x86, 4.9 is still the current release for the pipeline. Porting to 4.14/x86 requires an overhaul of the fpu management in the pipeline and cobalt cores (actually, the mainline changes requiring this were introduced in the 4.11 timeframe). The basic issue is about dropping the assumptions cobalt does about the kernel using the lazy fpu scheme, since mainline switched to strict eager mode for x86, dropping the former option entirely. > For 4.4, which we will need for product use for sure, I'm considering to > lift the support from 4.4 LTS to the CIP [1][2] kernel. That one will be > longer maintained and basically differs from LTS only in mainline > backports of certain hardware support and critical features. Plus some > reverts of broken LTS commits that will later on be reverted there as > well. The fuss that meltdown brings in would be a good point to do the > jump IMHO (there is no CIP kernel with meltdown fixes yet, due to issues > in LTS, but that will change soon). > No objection. > Regarding spectre, I'm not expected that much impact on ipipe, but that > situation is still too much in flux, even in upstream. > > What do you think? > I did not have any issue porting dovetail (4th gen pipeline) to arm/4.15; although the pipeline implementation is quite different, the mm portions are similar, so there is hope for the current pipeline with respect to mainline changes to the generic vm core, the situation with x86-specific bits is likely going to be different for obvious reasons. I did not check in details though. > @all: What are requirements of other users here? > > Jan > > [1] https://git.kernel.org/pub/scm/linux/kernel/git/bwh/linux-cip.git > [2] > https://wiki.linuxfoundation.org/civilinfrastructureplatform/cipkernelmaintenance > -- Philippe. ___ Xenomai mailing list Xenomai@xenomai.org https://xenomai.org/mailman/listinfo/xenomai
Re: [Xenomai] ipipe: kernel updates in the light of recent fixes (meltdown...)
I think using the CIP kernel would be a great idea. I can see this being popular with product developers. -Greg On Tue, Jan 23, 2018 at 12:52 PM, Jan Kiszka wrote: > Hi Philippe, > > as the number of question increase here, but I bet elsewhere as well, > and I'm planning to work on some parts soonish, I'd like to coordinate: > > Where are we standing with 4.14 right now? There is nothing for x86 yet, > right? > > What are the update plans for 4.9? We currently have one x86 user on > that kernel, but I suppose we could also move on to 4.14 once it's > ready. For me the question is right now where to invest x86-wise, 4.9 > update or 4.14 completion + update? > > For 4.4, which we will need for product use for sure, I'm considering to > lift the support from 4.4 LTS to the CIP [1][2] kernel. That one will be > longer maintained and basically differs from LTS only in mainline > backports of certain hardware support and critical features. Plus some > reverts of broken LTS commits that will later on be reverted there as > well. The fuss that meltdown brings in would be a good point to do the > jump IMHO (there is no CIP kernel with meltdown fixes yet, due to issues > in LTS, but that will change soon). > > Regarding spectre, I'm not expected that much impact on ipipe, but that > situation is still too much in flux, even in upstream. > > What do you think? > > @all: What are requirements of other users here? > > Jan > > [1] https://git.kernel.org/pub/scm/linux/kernel/git/bwh/linux-cip.git > [2] > https://wiki.linuxfoundation.org/civilinfrastructureplatform/cipkernelmaintenance > > -- > Siemens AG, Corporate Technology, CT RDA IOT SES-DE > Corporate Competence Center Embedded Linux > > ___ > Xenomai mailing list > Xenomai@xenomai.org > https://xenomai.org/mailman/listinfo/xenomai ___ Xenomai mailing list Xenomai@xenomai.org https://xenomai.org/mailman/listinfo/xenomai
Re: [Xenomai] [RFC] RTnet fixes and other updates
Philippe, On 23/01/18 18:21, Philippe Gerum wrote: > For review by interested parties, proposed for inclusion in 3.0.6. 3.0.7 ? Leopold > > Thanks, > > === > > The following changes since commit 3a2fbc62b2dce64e959a49db9bcc049e6acc4e21: > > cobalt/rtdm: factor out iovec[] copy routines (2018-01-23 18:12:19 +0100) > > are available in the git repository at: > > git://xenomai.org/xenomai-3.git wip/rtnet-fixes > > for you to review changes up to df4f825d1d8ab1b7129af59d43b3f219f4ae5101: > > net/socket: forward private ioctl requests to NIC driver (2018-01-23 > 18:12:28 +0100) > > > Philippe Gerum (26): > net/cap: fix panic in rtcap_signal_handler() > net: wire up corectl interface > net/tcp: fix invalid reference in getsockopt() > net/socket: add get_arg/put_arg helpers > net/iovec: add copy iterators for iovec[] > net/packet: ioctl: remove direct references to user memory > net/tcp: ioctl: remove direct references to user memory > net/udp: ioctl: remove direct references to user memory > net/packet: recvmsg: remove direct references to user memory > net/packet: recvmsg: write back namelen only if name required > net/packet: sendmsg: remove direct references to user memory > net/udp: recvmsg: remove direct references to user memory > net/udp: recvmsg: write back namelen only if name required > net/udp: sendmsg: remove direct references to user memory > net/tcp: recvmsg: remove direct references to user memory > net/tcp: sendmsg: remove direct references to user memory > net: convert to rtdm_get_iov_flatlen() > net/iovec: drop useless kernel<-> iovec[] copy helpers > net/udp: ioctl: fix temp arg buffer type > net/packet: ioctl: remove direct references to user memory (2) > net/udp: recvmsg, ioctl: remove direct references to user memory (2) > net/packet: ioctl: remove direct references to user memory (3) > net/socket: ioctl: remove direct references to user memory (2) > net/socket: enforce secondary mode for SIOCETHTOOL > net/socket: align rtdev do_ioctl handler on the regular ndo_do_ioctl > net/socket: forward private ioctl requests to NIC driver > > kernel/drivers/net/addons/cap.c | 12 +--- > kernel/drivers/net/drivers/igb/igb_main.c | 7 ++--- > kernel/drivers/net/stack/Makefile | 3 +- > kernel/drivers/net/stack/corectl.c | 9 -- > kernel/drivers/net/stack/include/rtdev.h| 2 +- > kernel/drivers/net/stack/include/rtnet_iovec.h | 26 +--- > kernel/drivers/net/stack/include/rtnet_socket.h | 10 -- > kernel/drivers/net/stack/iovec.c| 107 > +++ > kernel/drivers/net/stack/ipv4/tcp/tcp.c | 168 > +++ > kernel/drivers/net/stack/ipv4/udp/udp.c | 379 > +++- > kernel/drivers/net/stack/packet/af_packet.c | 334 > - > kernel/drivers/net/stack/rtnet_module.c | 6 > kernel/drivers/net/stack/socket.c | 230 > > 13 files changed, 823 insertions(+), 470 deletions(-) > -- -- Linux User 152692 GPG: 05F4A7A949A2D9AA Catalonia - A: Because it messes up the order in which people normally read text. Q: Why is top-posting such a bad thing? A: Top-posting. Q: What is the most annoying thing in e-mail? ___ Xenomai mailing list Xenomai@xenomai.org https://xenomai.org/mailman/listinfo/xenomai
