Hi Al Viro,
Today's next kernel build failed with commit 41e83b9: fix "netfilter:
xt_bpf: Fix XT_BPF_MODE_FD_PINNED
Machine Type: Power 7
kernel version: 4.15.0-rc2-next-20171204
kernel config: attached
test: 'make modules'
ERROR: ".bpf_prog_get_type_path" [net/netfilter/xt_bpf.ko] undefined!
Kernel built fine when bad commit is reverted.
--
Regard's
Abdul Haleem
IBM Linux Technology Centre
On Tue, Dec 05, 2017 at 08:13:11PM +0530, Abdul Haleem wrote:
> Hi Al Viro,
>
> Today's next kernel build failed with commit 41e83b9: fix "netfilter:
> xt_bpf: Fix XT_BPF_MODE_FD_PINNED
>
> Machine Type: Power 7
> kernel version: 4.15.0-rc2-next-20171204
> kernel config: attached
> test: 'make modules'
>
> ERROR: ".bpf_prog_get_type_path" [net/netfilter/xt_bpf.ko] undefined!
>
> Kernel built fine when bad commit is reverted.
In the same commit there's
+EXPORT_SYMBOL(bpf_prog_get_type_path);
so I really wonder WTF is going on...
Oh, I see. !BPF_SYSCALL and NETFILTER_XT_MATCH_BPF in .config triggers
that. Incremental (to be folded into the commit in question) follows:
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index a7000e4775e7..571450f98006 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -419,6 +419,8 @@ static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
attr->numa_node : NUMA_NO_NODE;
}
+struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
+
#else /* !CONFIG_BPF_SYSCALL */
static inline struct bpf_prog *bpf_prog_get(u32 ufd)
{
@@ -506,6 +508,12 @@ static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
{
return 0;
}
+
+static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
+ enum bpf_prog_type type)
+{
+ return -EOPNOTSUPP;
+}
#endif /* CONFIG_BPF_SYSCALL */
static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
@@ -514,7 +522,6 @@ static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
return bpf_prog_get_type_dev(ufd, type, false);
}
-struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
int bpf_prog_offload_compile(struct bpf_prog *prog);
On Tue, 2017-12-05 at 18:06 +0000, Al Viro wrote:
> On Tue, Dec 05, 2017 at 08:13:11PM +0530, Abdul Haleem wrote:
> > Hi Al Viro,
> >
> > Today's next kernel build failed with commit 41e83b9: fix "netfilter:
> > xt_bpf: Fix XT_BPF_MODE_FD_PINNED
> >
> > Machine Type: Power 7
> > kernel version: 4.15.0-rc2-next-20171204
> > kernel config: attached
> > test: 'make modules'
> >
> > ERROR: ".bpf_prog_get_type_path" [net/netfilter/xt_bpf.ko] undefined!
> >
> > Kernel built fine when bad commit is reverted.
>
> In the same commit there's
>
> +EXPORT_SYMBOL(bpf_prog_get_type_path);
>
> so I really wonder WTF is going on...
>
> Oh, I see. !BPF_SYSCALL and NETFILTER_XT_MATCH_BPF in .config triggers
> that. Incremental (to be folded into the commit in question) follows:
Thank you for the patch, the build succeeded. But I see some new
warnings.
./include/linux/bpf.h: In function ‘bpf_prog_get_type_path’:
./include/linux/bpf.h:515:2: warning: return makes pointer from integer
without a cast [enabled by default]
return -EOPNOTSUPP;
^
--
Regard's
Abdul Haleem
IBM Linux Technology Centre