On 1/15/2018 7:10 PM, Jason Gunthorpe wrote:
On Mon, Jan 15, 2018 at 07:01:13PM +0200, Yishai Hadas wrote:
On 1/12/2018 12:13 AM, Jason Gunthorpe wrote:
From: Jason Gunthorpe <jgg@xxxxxxxxxxxx>
To let us use definitions from the latest kernel headers in followup
patches.
From v4.15-rc7
Signed-off-by: Jason Gunthorpe <jgg@xxxxxxxxxxxx>
buildlib/RDMA_LinuxHeaders.cmake | 2 +-
buildlib/fixup-include/rdma-ib_user_verbs.h | 1176 +++++++++++++++++++++++++++
2 files changed, 1177 insertions(+), 1 deletion(-)
create mode 100644 buildlib/fixup-include/rdma-ib_user_verbs.h
diff --git a/buildlib/RDMA_LinuxHeaders.cmake b/buildlib/RDMA_LinuxHeaders.cmake
index ea60b9b55844ea..aa2a89a53d0c2d 100644
+++ b/buildlib/RDMA_LinuxHeaders.cmake
@@ -86,7 +86,7 @@ function(rdma_report_missing_kheaders)
endfunction()
# This list is topologically sorted
-rdma_check_kheader("rdma/ib_user_verbs.h" "${DEFAULT_TEST}" NO_SHIM OPTIONAL)
+rdma_check_kheader("rdma/ib_user_verbs.h" "int main(int argc,const char *argv[]) { return IB_USER_VERBS_EX_CMD_MODIFY_CQ; }")
Any future change in this kernel header will require a change in this check
as well and also a commit into fixup-include/rdma-ib_user_verbs.h with the
new kernel code, correct ?
Yes.
How will people be aware of that and won't miss it ? can we have some script
that will warn/suggest the expected change upon a kernel change ?
Keeping the header up to date is not a big problem, there is no way to
get new symbols otherwise so people patching in new features have to
do it.
travis always does compiles using these headers as well.
Keeping the check quoted above broadly correct is more difficult. I
can't think of how to automate that, other than to manually inspect
patches that update the header to make sure they update the check as
well.
We could also switch to a more iproute like model where we just
unconditionally use our copies of the kernel headers..
Are you referring to drop the cmake linux header check and always use a
private copy of the latest kernel UAPI file in rdma-core ? it can make
more sense comparing to current candidate code.
It can reside in a folder named 'rdma' and not installed by rdma-core as
applications don't use this file. If so we can drop the fixup file and
its matching check.
What do you think ?
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
