In a previous commit (1), BPF preload process was switched from user mode
process to use in-kernel light skeleton instead. However, in the kernel context
the available FD starts from 0, instead of normally 3 for user mode process.
The preload process also left two FDs open, taking over FD 0 and 1. This later
caused issues when kernel trys to setup stdin/stdout/stderr for init process,
assuming FD 0,1,2 are available.
As seen here:
Before fix:
ls -lah /proc/1/fd/*
lrwx------1 root root 64 Feb 23 17:20 /proc/1/fd/0 -> /dev/null
lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/1 -> /dev/null
lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/2 -> /dev/console
lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/6 -> /dev/console
lrwx------ 1 root root 64 Feb 23 17:20 /proc/1/fd/7 -> /dev/console
After Fix / Normal:
ls -lah /proc/1/fd/*
lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/0 -> /dev/console
lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/1 -> /dev/console
lrwx------ 1 root root 64 Feb 24 21:23 /proc/1/fd/2 -> /dev/console
In this patch:
- skel_closenz was changed to skel_closegez to correctly handle
FD=0 case.
- various places detecting FD > 0 was changed to FD >= 0.
- Call iterators_skel__detach() funciton to release FDs after links
are obtained.
1: commit cb80ddc ("bpf: Convert bpf_preload.ko to use light skeleton.")
Fixes: commit cb80ddc ("bpf: Convert bpf_preload.ko to use light skeleton.")
Signed-off-by: Yucong Sun <fallentree@xxxxxx>
V3 -> V1: removed all changes related to handle fd=0.
V2 -> V1: rename skel_closenez to skel_closegez, added comment as
requested.
---
kernel/bpf/preload/bpf_preload_kern.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/kernel/bpf/preload/bpf_preload_kern.c b/kernel/bpf/preload/bpf_preload_kern.c
index 30207c048d36..13cd0d146dd7 100644
--- a/kernel/bpf/preload/bpf_preload_kern.c
+++ b/kernel/bpf/preload/bpf_preload_kern.c
@@ -54,6 +54,16 @@ static int load_skel(void)
err = PTR_ERR(progs_link);
goto out;
}
+ /* Avoid taking over stdin/stdout/stderr of init process. This also
+ makes skel_closenz() no-op later in free_links_and_skel(). */
+ if (skel->links.dump_bpf_map_fd < 3) {
+ close_fd(skel->links.dump_bpf_map_fd);
+ skel->links.dump_bpf_map_fd = 0;
+ }
+ if (skel->links.dump_bpf_prog_fd < 3) {
+ close_fd(skel->links.dump_bpf_prog_fd);
+ skel->links.dump_bpf_prog_fd = 0;
+ }
return 0;
out:
free_links_and_skel();
--
2.30.2
