exec() fails if executed with SGX and the target's code segment is bigger than 4096 bytes about graphene HOT 3 CLOSED

This can also be triggered by running any (non-builtin) command via bash, for example:

LibOS/shim/test/apps/bash$ ./bash.manifest.sgx -c ls
[...]
[    1] loading file:/bin/ls as runtime object at 0x0
[    1] bkeep_mmap: 0x400000-0x41e000
[    1] memory fault at 0x0 (IP = 0xbd98e83)
[    1] SIGSEGV handled
[    1] run signal handler 0xbac1725 (11, 0xaf12650, 0xaf126d8)
[    1] killed by SIGSEGV
[...]

(I set loader.debug_type = inline in bash.manifest.template to get this output)

from graphene.

I have a same issue.
when I try to map 4096 bytes of memory with mmap function, the segmentation fault is triggered.
the output is following.
is this issue fixed?

[ 1] bkeep_mprotect: 0x5c4d3000-0x5c4d5000
[ 1] bkeep_munmap: 0x5c4d3000-0x5c4d5000
[ 1] ---- return from shim_munmap(...) = 0
[ 1] memory fault at 0x0 (IP = 0x5c553e06)
[ 1] SIGSEGV handled
[ 1] run signal handler 0x7bac5a4a (11, 0x64aaaa90, 0x64aaab18)
[ 1] killed by SIGSEGV

from graphene.

@adombeck I can no longer reproduce this issue (after the fixes in #215 ). I am assuming this is fixed in the latest code, unless you are still having issues.

@leesera can you share code to reproduce your issue? The log you list shows a segfault after an munmap, not on an mmap, which makes me suspect this is a different issue.

from graphene.