Re: port-sparc64/57219: ./configure for gcl-2.6.14 crashes kernel (reboot) on NetBSD 9.3 sparc64

[Alexander Schreiber <als%thangorodrim.ch@localhost>]

The following reply was made to PR port-sparc64/57219; it has been noted by GNATS.

From: Alexander Schreiber <als%thangorodrim.ch@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: port-sparc64/57219: ./configure for gcl-2.6.14 crashes kernel
 (reboot) on NetBSD 9.3 sparc64
Date: Thu, 9 Feb 2023 00:42:48 +0100

 Applied the aforementioned patch to the 9.3 release tree, kernel
 was rebuilt with the patch + DIAGNOSTIC + DEBUG and rebooted into
 it.
 
 Ran gcl ./configure again and it no longer crashes, but behaves
 as on amd64 (minus the segfault that doesn't happen there):
 
 configure: trying to adjust text start
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Abort trap              (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 [1]   Segmentation fault (core dumped) (${CC} ${LDFLAGS} -Wl,-T gcl.script foo.c -o f...
 configure: min log text start 19
 
 The console, however is pretty noisy, but that is expected when:
  - running with DEBUG
  - having user space code doing creative fuckery with VM mappings
    (as Lisps are prone to do)
 so there is lots of this:
 
 login: [ 1309.3109110] vmcmds 0/8, error 12
 [ 1309.3561572] vmcmd[0] = vmcmd_map_pagedvn 0/0x2000 fd@0 prot=05 flags=4
 [ 1309.4351499]      ^--- failed
 [ 1309.4705003] vmcmd[1] = vmcmd_map_readvn 0x100000/0x1388 fd@0 prot=07 flags=4
 [ 1309.5557414] vmcmd[2] = vmcmd_map_pagedvn 0x40200000/0x14000 fd@0 prot=05 flags=2
 [ 1309.6451453] vmcmd[3] = vmcmd_map_zero 0x14000/0x100000 fd@0 prot=00 flags=1
 [ 1309.7293514] vmcmd[4] = vmcmd_map_readvn 0x114000/0xe88 fd@0x14000 prot=03 flags=1
 [ 1309.8198050] vmcmd[5] = vmcmd_map_zero 0xfffffffff7efe000/0x100000 fd@0 prot=00 flags=0
 [ 1309.9154381] vmcmd[6] = vmcmd_map_zero 0xfffffffff7ffe000/0x7e00000 fd@0 prot=00 flags=8
 [ 1310.0121185] vmcmd[7] = vmcmd_map_zero 0xffffffffffdfe000/0x200000 fd@0 prot=03 flags=8
 [ 1310.1077601] execve_dovmcmds: vmcmd 0 failed: 12
 [ 1311.4319565] vmcmds 0/8, error 12
 [ 1311.4733745] vmcmd[0] = vmcmd_map_pagedvn 0/0x2000 fd@0 prot=05 flags=4
 [ 1311.5523666]      ^--- failed
 [ 1311.5877125] vmcmd[1] = vmcmd_map_readvn 0x100000/0x1388 fd@0 prot=07 flags=4
 [ 1311.6729583] vmcmd[2] = vmcmd_map_pagedvn 0x40200000/0x14000 fd@0 prot=05 flags=2
 [ 1311.7623631] vmcmd[3] = vmcmd_map_zero 0x14000/0x100000 fd@0 prot=00 flags=1
 [ 1311.8465705] vmcmd[4] = vmcmd_map_readvn 0x114000/0xe88 fd@0x14000 prot=03 flags=1
 [ 1311.9370176] vmcmd[5] = vmcmd_map_zero 0xfffffffff7efe000/0x100000 fd@0 prot=00 flags=0
 [ 1312.0326541] vmcmd[6] = vmcmd_map_zero 0xfffffffff7ffe000/0x7e00000 fd@0 prot=00 flags=8
 [ 1312.1293354] vmcmd[7] = vmcmd_map_zero 0xffffffffffdfe000/0x200000 fd@0 prot=03 flags=8
 [ 1312.2249828] execve_dovmcmds: vmcmd 0 failed: 12
 
 
 In summary: calling this fixed on actual sparc64 hardware as well.