BUG in pahole?: strange error in tests/flexible_arrays.sh
Hello all,
I ran into the following running the latest pahole tests:
$ vmlinux=~/linux/vmlinux ./tests/tests
1: Validation of BTF encoding of functions; this may take some time: Ok
2: Default BTF on a system without BTF: Ok
3: Flexible arrays accounting: pahole: type 'bpf_empty_prog_array' not found
pahole: type 'kstatmount' not found
pahole: type 'crypto_lskcipher' not found
pahole: type 'crypto_sig' not found
pahole: type 'hash_ctx' not found
pahole: type 'scsi_stream_status_header' not found
pahole: type 'virtnet_info' not found
pahole: type 'geneve_dev' not found
pahole: type 'geneve_config' not found
pahole: type 'lirc_fh' not found
pahole: type 'scmi_registered_events_desc' not found
pahole: type 'events_queue' not found
pahole: type 'hid_debug_list' not found
pahole: type 'flow_offload_action' not found
pahole: type 'nft_rule_dp_last' not found
pahole: type 'nft_rhash_elem' not found
pahole: type 'nft_hash_elem' not found
pahole: type 'nft_bitmap_elem' not found
pahole: type 'nft_rbtree_elem' not found
pahole: type 'nft_pipapo_elem' not found
pahole: type 'xt_standard_target' not found
pahole: type 'xt_error_target' not found
pahole: type 'ipt_standard' not found
pahole: type 'ipt_error' not found
pahole: type 'ip6t_standard' not found
pahole: type 'ip6t_error' not foundThis is simple to reproduce (e.g. code in flexible_arrays.sh):
$ pahole kstatmount ~/linux/vmlinux pahole: type 'kstatmount' not found
But despite the above:
$ bpftool btf dump file ~/linux/vmlinux format raw|grep kstatmount [13145] STRUCT 'kstatmount' size=624 vlen=8
And:
$ pahole -C kstatmount ~/linux/vmlinux
struct kstatmount {
struct statmount * buf; /* 0 4 */
...
};Has this been seen before? Am I missing something? Any insight folks have would be appreciated.
The same behaviour is also seen with '.tmp_vmlinux1.btf.o', which I attach.
Many thanks,
Tony Ambardar
Quoted 52 lines · Tony Ambardar
Hello all,
I ran into the following running the latest pahole tests:
$ vmlinux=~/linux/vmlinux ./tests/tests 1: Validation of BTF encoding of functions; this may take some time: Ok 2: Default BTF on a system without BTF: Ok 3: Flexible arrays accounting: pahole: type 'bpf_empty_prog_array' not found pahole: type 'kstatmount' not found pahole: type 'crypto_lskcipher' not found pahole: type 'crypto_sig' not found pahole: type 'hash_ctx' not found pahole: type 'scsi_stream_status_header' not found pahole: type 'virtnet_info' not found pahole: type 'geneve_dev' not found pahole: type 'geneve_config' not found pahole: type 'lirc_fh' not found pahole: type 'scmi_registered_events_desc' not found pahole: type 'events_queue' not found pahole: type 'hid_debug_list' not found pahole: type 'flow_offload_action' not found pahole: type 'nft_rule_dp_last' not found pahole: type 'nft_rhash_elem' not found pahole: type 'nft_hash_elem' not found pahole: type 'nft_bitmap_elem' not found pahole: type 'nft_rbtree_elem' not found pahole: type 'nft_pipapo_elem' not found pahole: type 'xt_standard_target' not found pahole: type 'xt_error_target' not found pahole: type 'ipt_standard' not found pahole: type 'ipt_error' not found pahole: type 'ip6t_standard' not found pahole: type 'ip6t_error' not foundThis is simple to reproduce (e.g. code in
flexible_arrays.sh):$ pahole kstatmount ~/linux/vmlinux pahole: type 'kstatmount' not foundBut despite the above:
$ bpftool btf dump file ~/linux/vmlinux format raw|grep kstatmount [13145] STRUCT 'kstatmount' size=624 vlen=8And:
$ pahole -C kstatmount ~/linux/vmlinux struct kstatmount { struct statmount * buf; /* 0 4 */ ... };
Quoted 7 lines
Has this been seen before? Am I missing something? Any insight folks have would be appreciated.
The same behaviour is also seen with '.tmp_vmlinux1.btf.o', which I attach.
Many thanks,
Tony Ambardar
hi Tony, I've seen this too (with a slightly different cast of characters); I keep meaning to look into it but haven't had a chance yet so thanks for doing some investigation! Seems like it is a bug in type display rather than in BTF generation at least..
Alan
Quoted 61 lines · Alan Maguire
Quoted 52 lines · Tony Ambardar
Hello all,
I ran into the following running the latest pahole tests:
$ vmlinux=~/linux/vmlinux ./tests/tests 1: Validation of BTF encoding of functions; this may take some time: Ok 2: Default BTF on a system without BTF: Ok 3: Flexible arrays accounting: pahole: type 'bpf_empty_prog_array' not found pahole: type 'kstatmount' not found pahole: type 'crypto_lskcipher' not found pahole: type 'crypto_sig' not found pahole: type 'hash_ctx' not found pahole: type 'scsi_stream_status_header' not found pahole: type 'virtnet_info' not found pahole: type 'geneve_dev' not found pahole: type 'geneve_config' not found pahole: type 'lirc_fh' not found pahole: type 'scmi_registered_events_desc' not found pahole: type 'events_queue' not found pahole: type 'hid_debug_list' not found pahole: type 'flow_offload_action' not found pahole: type 'nft_rule_dp_last' not found pahole: type 'nft_rhash_elem' not found pahole: type 'nft_hash_elem' not found pahole: type 'nft_bitmap_elem' not found pahole: type 'nft_rbtree_elem' not found pahole: type 'nft_pipapo_elem' not found pahole: type 'xt_standard_target' not found pahole: type 'xt_error_target' not found pahole: type 'ipt_standard' not found pahole: type 'ipt_error' not found pahole: type 'ip6t_standard' not found pahole: type 'ip6t_error' not foundThis is simple to reproduce (e.g. code in
flexible_arrays.sh):$ pahole kstatmount ~/linux/vmlinux pahole: type 'kstatmount' not foundBut despite the above:
$ bpftool btf dump file ~/linux/vmlinux format raw|grep kstatmount [13145] STRUCT 'kstatmount' size=624 vlen=8And:
$ pahole -C kstatmount ~/linux/vmlinux struct kstatmount { struct statmount * buf; /* 0 4 */ ... };Quoted 7 lines
Has this been seen before? Am I missing something? Any insight folks have would be appreciated.
The same behaviour is also seen with '.tmp_vmlinux1.btf.o', which I attach.
Many thanks,
Tony Ambardar
Quoted 6 lines
hi Tony, I've seen this too (with a slightly different cast of characters); I keep meaning to look into it but haven't had a chance yet so thanks for doing some investigation! Seems like it is a bug in type display rather than in BTF generation at least..
Alan
Thanks for confirming this -- I was worried it might be a side-effect of my function-encoding patch. Out of curiosity, how does option "-C" affect BTF parsing or pahole operation? It wasn't clear to me just what "Show just this class" means/implies...
Cheers,
Tony