Segfault due to unchecked return value

[smithzvk]

jna/native/dispatch.c

Line 3461 in dabdcd7

closure = ffi_closure_alloc(sizeof(ffi_closure), &code);

I think I traced a segfault to this line. The return type stored in closure can be NULL on allocation failure. ffi_prep_closure_loc assumes that the value is valid, so passing NULL will produce a seg fault.

This happens on locked down systems with restricted execution and write permissions. The libffi closure API requires either trivially executable heap allocations or a location to mmap with both read-write and exec permissions. If it doesn't have this (e.g. selinux with no rw, exec mounts), it will fail to allocate the closure structure.

Perhaps this could fail more gracefully?

[twall]

There's not currently any explicit checking for whether it's _possible_ to execute callbacks/allocate executable memory, so to the extent that we can throw an exception rather than segfault, that would be nice. PRs welcome.

…

On Thu, Jun 13, 2019 at 10:30 AM Zach Kost-Smith ***@***.***> wrote: https://github.com/java-native-access/jna/blob/dabdcd7c27e2b4d7f672fca1c55b6da5c74badb5/native/dispatch.c#L3461 I think I traced a segfault to this line. The return type stored in closure can be NULL on allocation failure. ffi_prep_closure_loc assumes that the value is valid, so passing NULL will produce a seg fault. This happens on locked down systems with restricted execution and write permissions. The libffi closure API requires either trivially executable heap allocations or a location to mmap with both read-write and exec permissions. If it doesn't have this (e.g. selinux with no rw, exec mounts), it will fail to allocate the closure structure. Perhaps this could fail more gracefully? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1107?email_source=notifications&email_token=AAFYZLMJKYYGQZTJZZ7YFM3P2JKW5A5CNFSM4HX2GHLKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4GZKSZDQ>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAFYZLOLUXNQUXKHEOYGVDTP2JKW5ANCNFSM4HX2GHLA> .

[smithzvk]

I'll take a look as time permits. I don't think I can get my company to sponsor this work, unfortunately, so it may take a while for the free time to surface.