On Thu, 12 Oct 2017 06:21:30 -0400
Jiong Wang <jiong.wang@...> wrote:

We came across an llvm bug when compiling some testcases that 64-bit
immediates are silently truncated into 32-bit and then packed into
BPF_JMP | BPF_K encoding. This caused comparison with wrong value.

I think you send this to the wrong mailing list... this looks like a
patch against the LLVM source code.

Shouldn't you send to: llvm-dev@... ?

This bug looks to be introduced by r308080.

This looks like a very recent change:
https://llvm.org/viewvc/llvm-project?view=revision&revision=308080
Sat Jul 15 05:41:42 2017 UTC (2 months, 4 weeks ago)

(As you are sending this to a user mailing list: xdp-newbies@...)

I want to know if this have made it into a LLVM release? and which release?

As the git-repo[1] replica of LLVM SVN-repo does not git-tag the
releases, I could not answer this question myself via the command:

$ git describe --contains 7c423e0690

[1] http://llvm.org/git/llvm.git

The Select_Ri pattern is
supposed to be lowered into J*_Ri while the latter only support 32-bit
immediate encoding, therefore Select_Ri should have similar immediate
predicate check as what J*_Ri are doing.

Reported-by: Jakub Kicinski <jakub.kicinski@...>
Signed-off-by: Jiong Wang <jiong.wang@...>
---
lib/Target/BPF/BPFISelLowering.cpp | 8 ++++++--
lib/Target/BPF/BPFInstrInfo.td | 2 +-
test/CodeGen/BPF/select_ri.ll | 35 +++++++++++++++++++++++++++++++++++
3 files changed, 42 insertions(+), 3 deletions(-)

--
Best regards,
Jesper Dangaard Brouer
MSc.CS, Principal Kernel Engineer at Red Hat
LinkedIn: http://www.linkedin.com/in/brouer