Closed dmitriisilin closed 3 years ago
rhughes-xilinx
commented
3 years ago I agree that this is suboptimal, but it's ultimately harmless: there's no reason to pass any of the realtime signals to EF_SIGNALS_NOPOSTPONE because they already have they behaviour anyway.
If you'd like to propose a pull request to clean it up, I'd happily accept it.
dmitriisilin
commented
3 years ago I was trying to build project, to make sure my changes will pass unit tests before opening a merge request. However I'm experiencing build issues. Looks as kernel_compat.h needs some explicit casts. I'll try to workaround failures and see how it goes. In case, I discover more issues with build system, is there a "pre-integration pipeline" for the project?, In other words, can I propose a fix for the original problem without testing?
make[3]: Entering directory `/usr/src/kernels/3.10.0-229.el7.x86_64' CC /home/dsilin/ws/github_onload/onload/build/x86_64_linux-3.10.0-229.el7.x86_64/lib/cplane/mib.o CC /home/dsilin/ws/github_onload/onload/build/x86_64_linux-3.10.0-229.el7.x86_64/lib/cplane/mib_fwd.o CC /home/dsilin/ws/github_onload/onload/build/x86_64_linux-3.10.0-229.el7.x86_64/lib/cplane/services.o CC /home/dsilin/ws/github_onload/onload/build/x86_64_linux-3.10.0-229.el7.x86_64/lib/cplane/onload.o In file included from /home/dsilin/ws/github_onload/onload/build/x86_64_linux-3.10.0-229.el7.x86_64/lib/cplane/onload.c:8:0: /home/dsilin/ws/github_onload/onload/src/include/ci/driver/kernel_compat.h: In function \u2018pin_user_pages\u2019: /home/dsilin/ws/github_onload/onload/src/include/ci/driver/kernel_compat.h:163:42: error: passing argument 1 of \u2018get_user_pages\u2019 makes pointer from integer without a cast [-Werror] pages, vmas); ^ In file included from include/linux/scatterlist.h:6:0, from include/linux/dmaengine.h:27, from include/linux/skbuff.h:31, from include/linux/icmpv6.h:4, from include/linux/ipv6.h:59, from /home/dsilin/ws/github_onload/onload/src/include/ci/tools/platform/linux_kernel.h:54, from /home/dsilin/ws/github_onload/onload/src/include/ci/tools/sysdep.h:56, from /home/dsilin/ws/github_onload/onload/src/include/cplane/mib.h:25, from /home/dsilin/ws/github_onload/onload/src/include/cplane/cplane.h:21, from /home/dsilin/ws/github_onload/onload/build/x86_64_linux-3.10.0-229.el7.x86_64/lib/cplane/onload.c:4: include/linux/mm.h:1159:6: note: expected \u2018struct task_struct \u2019 but argument is of type \u2018long unsigned int\u2019 long get_user_pages(struct task_struct tsk, struct mm_struct *mm,
rhughes-xilinx
commented
3 years ago We no longer support RHEL7.1 (and Red Hat stopped supporting it 4 years ago).
ol-alexandra
commented
3 years ago I've created internal bug ON-13290. I'll fix it in the master branch and for upcoming onload-7.1.2.
@rhughes-xilinx
there's no reason to pass any of the realtime signals to EF_SIGNALS_NOPOSTPONE because they already have they behaviour anyway
Can you explain, please?
rhughes-xilinx
commented
3 years ago there's no reason to pass any of the realtime signals to EF_SIGNALS_NOPOSTPONE because they already have they behaviour anyway
Can you explain, please?
No, I can't: I misread the code.
ol-alexandra
commented
3 years ago Fixed in master and onload-7.1. If we ever release an update for EOL6, it will be included there as well.
src/lib/transport/ip/signal.c
if (our_info && our_info->inside_lib && (CITP_OPTS.signals_no_postpone & (1 << (signum-1))) == 0)4 bytes integer instead of 8 bytes is getting shifted