On Sat, Mar 01, 2025 at 01:38:16PM +0100, Arnd Bergmann wrote:
> On Sat, Mar 1, 2025, at 13:28, Stanislaw Gruszka wrote:
> > On Tue, Feb 25, 2025 at 03:53:53PM +0100, Arnd Bergmann wrote:
> >
> > But then the code will be compiled for !CONFIG_MAC80211_DEBUGFS
> > case, it does compile for me:
> >
> > - 22475 1160 0 23635
> > 5c53 drivers/net/wireless/intel/iwlegacy/4965-rs.o
> > + 23008 1168 0 24176
> > 5e70 drivers/net/wireless/intel/iwlegacy/4965-rs.o
>
> Very strange, this really shouldn't happen. Which symbols
> exactly do you see the compiler fail to drop with my patch,
nm 4965-rs.o diffrence before and after patch:
U ieee80211_rate_control_register
U ieee80211_rate_control_unregister
U ieee80211_start_tx_ba_session
U ieee80211_stop_tx_ba_session
0000000000000010 t il4965_hwrate_to_plcp_idx
-00000000000043b0 T il4965_rate_control_register
-00000000000043e0 T il4965_rate_control_unregister
+00000000000043f0 T il4965_rate_control_register
+0000000000004420 T il4965_rate_control_unregister
00000000000002e0 t il4965_rate_n_flags_from_tbl
0000000000000000 t il4965_rate_n_flags_from_tbl.cold
0000000000000270 t il4965_rs_alloc
0000000000000710 t il4965_rs_alloc_sta
00000000000000dd t il4965_rs_alloc_sta.cold
-0000000000001810 t il4965_rs_collect_tx_data.isra.0
-00000000000012b0 t il4965_rs_fill_link_cmd
-0000000000000495 t il4965_rs_fill_link_cmd.cold
+0000000000001850 t il4965_rs_collect_tx_data.isra.0
+0000000000000e90 t il4965_rs_dbgfs_set_mcs.isra.0
+00000000000002f6 t il4965_rs_dbgfs_set_mcs.isra.0.cold
+0000000000001340 t il4965_rs_fill_link_cmd
+0000000000000518 t il4965_rs_fill_link_cmd.cold
00000000000002a0 t il4965_rs_free
0000000000000e40 t il4965_rs_free_sta
000000000000028d t il4965_rs_free_sta.cold
@@ -173,31 +184,31 @@
0000000000000a20 t il4965_rs_get_best_rate
0000000000000bb0 t il4965_rs_get_rate
0000000000000209 t il4965_rs_get_rate.cold
-0000000000001180 t il4965_rs_get_tbl_info_from_mcs.isra.0
-0000000000003bd0 T il4965_rs_rate_init
-0000000000000d5f t il4965_rs_rate_init.cold
+0000000000001210 t il4965_rs_get_tbl_info_from_mcs.isra.0
+0000000000003c10 T il4965_rs_rate_init
+0000000000000de4 t il4965_rs_rate_init.cold
00000000000002c0 t il4965_rs_rate_init_stub
0000000000000760 t il4965_rs_set_expected_tpt_table
0000000000000490 t il4965_rs_stay_in_table
0000000000000088 t il4965_rs_stay_in_table.cold
-0000000000001000 t il4965_rs_switch_to_mimo2.isra.0
-00000000000003c3 t il4965_rs_switch_to_mimo2.isra.0.cold
-0000000000000e90 t il4965_rs_switch_to_siso.isra.0
-00000000000002f6 t il4965_rs_switch_to_siso.isra.0.cold
+0000000000001090 t il4965_rs_switch_to_mimo2.isra.0
+0000000000000446 t il4965_rs_switch_to_mimo2.isra.0.cold
+0000000000000f20 t il4965_rs_switch_to_siso.isra.0
+0000000000000379 t il4965_rs_switch_to_siso.isra.0.cold
00000000000000d0 t il4965_rs_tl_rm_old_stats
00000000000001a0 t il4965_rs_toggle_antenna
-0000000000001970 t il4965_rs_tx_status
-00000000000004d4 t il4965_rs_tx_status.cold
+00000000000019b0 t il4965_rs_tx_status
+0000000000000559 t il4965_rs_tx_status.cold
U il_debug_level
U il_is_ht40_tx_allowed
-0000000000001180 R il_rates
+00000000000012a0 R il_rates
U il_send_lq_cmd
U jiffies
U jiffies_to_msecs
000000000000004b r .LC15
0000000000000070 r .LC9
-0000000000001100 r rs_4965_ops
-0000000000001210 r rs_ht_to_legacy
+0000000000001220 r rs_4965_ops
+0000000000001330 r rs_ht_to_legacy
U __stack_chk_fail
U __ubsan_handle_out_of_bounds
U __ubsan_handle_shift_out_of_bounds
> and which compiler version are you using?
It is:
gcc (GCC) 9.3.1 20200408 (Red Hat 9.3.1-2)
I've checked on other system with
gcc (GCC) 14.2.1 20240912 (Red Hat 14.2.1-3)
and there size difference is similar:
- 28876 4875 0 33751 83d7 drivers/net/wireless/intel/iwlegacy/4965-rs.o
+ 29454 4851 0 34305 8601 drivers/net/wireless/intel/iwlegacy/4965-rs.o
> > How about moving
> > static const struct il_rate_mcs_info il_rate_mcs[RATE_COUNT]
> > under CONFIG_MAC80211_DEBUGFS ? Maybe inside the function that use it ?
>
> It's not supposed to make a difference, let's try to figure
> out if there is a compiler bug or a mistake in my patch first
> and then fix it in the right place.
Regards
Stanislaw
