Re: [PATCH V3 0/3] eBPF support

["Jose E. Marchesi" <jose.marchesi@oracle.com>]

Hi Daniel.

    On Fri Jul 3, 2020 at 5:46 AM PDT, Jose E. Marchesi via Gdb-patches wrote:
    > [Changes from V2:
    > - Urgh, V2 contained a patch still too big for mailman. In this
    > version I am splitting the simulator part in two commits. Thunks
    > for fully generated configure files are still removed.]
    >
    > Hi good peoples!
    >
    > This patch series adds support for the eBPF virtual architecture to
    > GDB [1].
    >
    > The first patch contains the basic bits to GDB in order to support the
    > bpf-unknown-none target. Breakpointing and instruction
    > single-stepping works, but the debugging support in eBPF is still very
    > minimal. This is mainly due to the many limitations imposed by the
    > architecture (disjoint stack, maximum stack size, etc). We are
    > working to overcome these limitations, by introducing a variant called
    > xbpf, already supported in GCC with the -mxbpf option, whose purpose
    > is to ease debugging and to be used in other contexts different than
    > the Linux kernel, less restrictive.
    >
    > The second patch adds a basic CGEN-based instruction simulator for
    > eBPF. It can run many eBPF programs and works well with GDB. A
    > testsuite covering the supported instructions is also included. We
    > will be expanding it in order to emulate the several kernel contexts
    > in which eBPF programs can run, so eBPF developers can use GDB to
    > debug their programs without having to load them in a running kernel.
    > Currently the only kernel helper implemented in the simulator is
    > printk, which is used by the tests.
    >
    > We of course commit to maintain and evolve this stuff :)
    >
    > [1] Support for eBPF has been already added to both binutils and GCC.
    
    This is quite exciting for bpf developers.

There is still a lot of work to do, but I am positive that eventually we
will achieve a good source-level debugging experience :)

    Are there any instructions on how to test this out? Here's what I've
    tried so far after I applied your patches:
    
        (gdb) file ~/dev/libbpf-rs/target/bpf/runqslower.bpf.o
        Reading symbols from ~/dev/libbpf-rs/target/bpf/runqslower.bpf.o...

Try with something like:

(gdb) target sim
(gdb) sim memory-size 4Mb
(gdb) load ~/dev/libbpf-rs/target/bpf/runqslower.bpf.o
(gdb) file load ~/dev/libbpf-rs/target/bpf/runqslower.bpf.o