Re: current issues on Octeon (ER-4)

[Denis Ovsienko <denis%ovsienko.info@localhost>]

On Sat, 20 May 2023 22:56:33 +0200 (CEST)
Havard Eidnes <he%NetBSD.org@localhost> wrote:

> n64: {6} file /usr/pkg/bin/git
> /usr/pkg/bin/git: ELF 64-bit MSB pie executable, MIPS, MIPS-III
> version 1 (SYSV), dynamically linked, interpreter
> /usr/libexec/ld.elf_so, for NetBSD 9.99.106, not stripped

Thank you for the detailed comments Havard.

For posterity, when I raised the dk1 problem last time, the workaround
was to give either "root=dk1" or "root=wedge:octeon-root" parameter to
NetBSD kernel from U-boot.

Your git executable seems to be for the evbmips-mipsn64eb (64-bit
userspace) variant.  I forgot to mention my test was on the
evbmips-mips64eb (32-bit userspace) variant:

# uname -a
NetBSD 202305140430 10.99.4 NetBSD 10.99.4 (OCTEON) #0: Sun May 14
01:17:08 UTC 2023
mkrepro%mkrepro.NetBSD.org@localhost:/usr/src/sys/arch/evbmips/compile/OCTEON
evbmips

# sysctl hw.machine_arch
hw.machine_arch = mips64eb

When I tried evbmips-mipsn64eb in July 2021, it just came into existence
and did not work well enough for my purposes.  Is it considered stable
now?

Anyway, I tried to get better diagnostics on what I use, and apparently
the git problem no longer reproduces after reinstalling the same
snapshot and the same previously compiled git package:

# file /usr/pkg/bin/git | tr ',' '\n'
/usr/pkg/bin/git: ELF 32-bit MSB pie executable
 MIPS
 N32 MIPS-III version 1 (SYSV)
 dynamically linked
 interpreter /usr/libexec/ld.elf_so
 for NetBSD 10.99.4
 not stripped

# git --version
git version 2.40.0

This stands for both the octeon.img.gz method and the
netbsd-INSTALL_OCTEON.gz method.  Out of interest I booted the
previously generated image again and there the problem reproduced, so I
took additional diagnostics and it looks as if the executable is
damaged:

# gdb /usr/pkg/bin/git
[...]
"0x7fff7770s": not in executable format: file format not recognized
(gdb) q

# file /usr/pkg/bin/git | tr ',' '\n'
/usr/pkg/bin/git: ELF 32-bit MSB pie executable
 MIPS
 N32 MIPS-III version 1 (SYSV)
 dynamically linked
 interpreter /usr/libexec/ld.elf_so
 missing section headers at 4635216

So it seems the root cause of the git issue was filesystem damage
during the OS crash, and the 202305140430 snapshot is usable enough to
upgrade the host and to see how it works.  Let me take it from there.

Before I forget, there is one other rough edge with evbmips-mips64eb:
* When booting octeon.img.gz, DHCP client works.
* When booting netbsd-OCTEON after the manual installation, DHCP client
  works.
* When booting netbsd-INSTALL_OCTEON, DHCP client does not work because
  the kernel does not seem to have BPF (this has been this way at least
  since 2021), also there is an assertion at the end.  Using static
  IPv4 configuration in the installer works around the problem.

     Status: Command failed
    Command: /sbin/dhcpcd -d -n cnmac1
     Hit enter to continue
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
Berkley Packet Filter not found
cnmac1: sending DISCOVER (xid 0xeb30a157), next in 8.4 seconds
cnmac1: sending DISCOVER (xid 0xeb30a157), next in 8.4 seconds
Berkley Packet Filter not found
Berkley Packet Filter not found
cnmac1: sending DISCOVER (xid 0xeb30a157), next in 15.6 seconds
cnmac1: sending DISCOVER (xid 0xeb30a157), next in 15.6 seconds
Berkley Packet Filter not found
Berkley Packet Filter not found
timed out
timed out
cnmac1: executing: /libexec/dhcpcd-run-hooks STOPPED
cnmac1: executing: /libexec/dhcpcd-run-hooks STOPPED
dhcpcd_fork_cb: truncated read 0 (expected 4)
dhcpcd_fork_cb: truncated read 0 (expected 4)
assertion "ifp != NULL" failed: file "/usr/src/lib/libc/net/getifaddrs.c", line
303, function "_freeifaddrs"

-- 
    Denis Ovsienko