2023-03-23 08:32:31

by Ping-Ke Shih

[permalink] [raw]
Subject: [PATCH] wifi: rtw89: fix potential race condition between napi_init and napi_enable

A race condition can happen if netdev is registered, but NAPI isn't
initialized yet, and meanwhile user space starts the netdev that will
enable NAPI. Then, it hits BUG_ON():

kernel BUG at net/core/dev.c:6423!
invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
CPU: 0 PID: 417 Comm: iwd Not tainted 6.2.7-slab-dirty #3 eb0f5a8a9d91
Hardware name: LENOVO 21DL/LNVNB161216, BIOS JPCN20WW(V1.06) 09/20/2022
RIP: 0010:napi_enable+0x3f/0x50
Code: 48 89 c2 48 83 e2 f6 f6 81 89 08 00 00 02 74 0d 48 83 ...
RSP: 0018:ffffada1414f3548 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffffa01425802080 RCX: 0000000000000000
RDX: 00000000000002ff RSI: ffffada14e50c614 RDI: ffffa01425808dc0
RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000001 R11: 0000000000000100 R12: ffffa01425808f58
R13: 0000000000000000 R14: ffffa01423498940 R15: 0000000000000001
FS: 00007f5577c0a740(0000) GS:ffffa0169fc00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f5577a19972 CR3: 0000000125a7a000 CR4: 0000000000750ef0
PKRU: 55555554
Call Trace:
<TASK>
rtw89_pci_ops_start+0x1c/0x70 [rtw89_pci 6cbc75429515c181cbc386478d5cfb32ffc5a0f8]
rtw89_core_start+0xbe/0x160 [rtw89_core fe07ecb874820b6d778370d4acb6ef8a37847f22]
rtw89_ops_start+0x26/0x40 [rtw89_core fe07ecb874820b6d778370d4acb6ef8a37847f22]
drv_start+0x42/0x100 [mac80211 c07fa22af8c3cf3f7d7ab3884ca990784d72e2d2]
ieee80211_do_open+0x311/0x7d0 [mac80211 c07fa22af8c3cf3f7d7ab3884ca990784d72e2d2]
ieee80211_open+0x6a/0x90 [mac80211 c07fa22af8c3cf3f7d7ab3884ca990784d72e2d2]
__dev_open+0xe0/0x180
__dev_change_flags+0x1da/0x250
dev_change_flags+0x26/0x70
do_setlink+0x37c/0x12c0
? ep_poll_callback+0x246/0x290
? __nla_validate_parse+0x61/0xd00
? __wake_up_common_lock+0x8f/0xd0

To fix this, follow Jonas' suggestion to switch the order of these
functions and move register netdev to be the last step of PCI probe.
Also, correct the error handling of rtw89_core_register_hw().

Fixes: e3ec7017f6a2 ("rtw89: add Realtek 802.11ax driver")
Cc: [email protected]
Reported-by: Hyeonggon Yoo <[email protected]>
Link: https://lore.kernel.org/linux-wireless/CAOiHx=n7EwK2B9CnBR07FVA=sEzFagb8TkS4XC_qBNq8OwcYUg@mail.gmail.com/T/#t
Suggested-by: Jonas Gorski <[email protected]>
Tested-by: Larry Finger<[email protected]>
Reviewed-by: Larry Finger<[email protected]>
Signed-off-by: Ping-Ke Shih <[email protected]>
---
drivers/net/wireless/realtek/rtw89/core.c | 10 +++++++---
drivers/net/wireless/realtek/rtw89/pci.c | 19 ++++++++++---------
2 files changed, 17 insertions(+), 12 deletions(-)

diff --git a/drivers/net/wireless/realtek/rtw89/core.c b/drivers/net/wireless/realtek/rtw89/core.c
index 489fa7a86160d..1efe008706669 100644
--- a/drivers/net/wireless/realtek/rtw89/core.c
+++ b/drivers/net/wireless/realtek/rtw89/core.c
@@ -3434,18 +3434,22 @@ static int rtw89_core_register_hw(struct rtw89_dev *rtwdev)
ret = ieee80211_register_hw(hw);
if (ret) {
rtw89_err(rtwdev, "failed to register hw\n");
- goto err;
+ goto err_free_supported_band;
}

ret = rtw89_regd_init(rtwdev, rtw89_regd_notifier);
if (ret) {
rtw89_err(rtwdev, "failed to init regd\n");
- goto err;
+ goto err_unregister_hw;
}

return 0;

-err:
+err_unregister_hw:
+ ieee80211_unregister_hw(hw);
+err_free_supported_band:
+ rtw89_core_clr_supported_band(rtwdev);
+
return ret;
}

diff --git a/drivers/net/wireless/realtek/rtw89/pci.c b/drivers/net/wireless/realtek/rtw89/pci.c
index ec8bb5f10482e..4025aabe06042 100644
--- a/drivers/net/wireless/realtek/rtw89/pci.c
+++ b/drivers/net/wireless/realtek/rtw89/pci.c
@@ -3874,25 +3874,26 @@ int rtw89_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
rtw89_pci_link_cfg(rtwdev);
rtw89_pci_l1ss_cfg(rtwdev);

- ret = rtw89_core_register(rtwdev);
- if (ret) {
- rtw89_err(rtwdev, "failed to register core\n");
- goto err_clear_resource;
- }
-
rtw89_core_napi_init(rtwdev);

ret = rtw89_pci_request_irq(rtwdev, pdev);
if (ret) {
rtw89_err(rtwdev, "failed to request pci irq\n");
- goto err_unregister;
+ goto err_deinit_napi;
+ }
+
+ ret = rtw89_core_register(rtwdev);
+ if (ret) {
+ rtw89_err(rtwdev, "failed to register core\n");
+ goto err_free_irq;
}

return 0;

-err_unregister:
+err_free_irq:
+ rtw89_pci_free_irq(rtwdev, pdev);
+err_deinit_napi:
rtw89_core_napi_deinit(rtwdev);
- rtw89_core_unregister(rtwdev);
err_clear_resource:
rtw89_pci_clear_resource(rtwdev, pdev);
err_declaim_pci:
--
2.25.1


2023-04-03 13:47:06

by Kalle Valo

[permalink] [raw]
Subject: Re: [PATCH] wifi: rtw89: fix potential race condition between napi_init and napi_enable

Ping-Ke Shih <[email protected]> wrote:

> A race condition can happen if netdev is registered, but NAPI isn't
> initialized yet, and meanwhile user space starts the netdev that will
> enable NAPI. Then, it hits BUG_ON():
>
> kernel BUG at net/core/dev.c:6423!
> invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
> CPU: 0 PID: 417 Comm: iwd Not tainted 6.2.7-slab-dirty #3 eb0f5a8a9d91
> Hardware name: LENOVO 21DL/LNVNB161216, BIOS JPCN20WW(V1.06) 09/20/2022
> RIP: 0010:napi_enable+0x3f/0x50
> Code: 48 89 c2 48 83 e2 f6 f6 81 89 08 00 00 02 74 0d 48 83 ...
> RSP: 0018:ffffada1414f3548 EFLAGS: 00010246
> RAX: 0000000000000000 RBX: ffffa01425802080 RCX: 0000000000000000
> RDX: 00000000000002ff RSI: ffffada14e50c614 RDI: ffffa01425808dc0
> RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000000000000
> R10: 0000000000000001 R11: 0000000000000100 R12: ffffa01425808f58
> R13: 0000000000000000 R14: ffffa01423498940 R15: 0000000000000001
> FS: 00007f5577c0a740(0000) GS:ffffa0169fc00000(0000) knlGS:0000000000000000
> CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 00007f5577a19972 CR3: 0000000125a7a000 CR4: 0000000000750ef0
> PKRU: 55555554
> Call Trace:
> <TASK>
> rtw89_pci_ops_start+0x1c/0x70 [rtw89_pci 6cbc75429515c181cbc386478d5cfb32ffc5a0f8]
> rtw89_core_start+0xbe/0x160 [rtw89_core fe07ecb874820b6d778370d4acb6ef8a37847f22]
> rtw89_ops_start+0x26/0x40 [rtw89_core fe07ecb874820b6d778370d4acb6ef8a37847f22]
> drv_start+0x42/0x100 [mac80211 c07fa22af8c3cf3f7d7ab3884ca990784d72e2d2]
> ieee80211_do_open+0x311/0x7d0 [mac80211 c07fa22af8c3cf3f7d7ab3884ca990784d72e2d2]
> ieee80211_open+0x6a/0x90 [mac80211 c07fa22af8c3cf3f7d7ab3884ca990784d72e2d2]
> __dev_open+0xe0/0x180
> __dev_change_flags+0x1da/0x250
> dev_change_flags+0x26/0x70
> do_setlink+0x37c/0x12c0
> ? ep_poll_callback+0x246/0x290
> ? __nla_validate_parse+0x61/0xd00
> ? __wake_up_common_lock+0x8f/0xd0
>
> To fix this, follow Jonas' suggestion to switch the order of these
> functions and move register netdev to be the last step of PCI probe.
> Also, correct the error handling of rtw89_core_register_hw().
>
> Fixes: e3ec7017f6a2 ("rtw89: add Realtek 802.11ax driver")
> Cc: [email protected]
> Reported-by: Hyeonggon Yoo <[email protected]>
> Link: https://lore.kernel.org/linux-wireless/CAOiHx=n7EwK2B9CnBR07FVA=sEzFagb8TkS4XC_qBNq8OwcYUg@mail.gmail.com/T/#t
> Suggested-by: Jonas Gorski <[email protected]>
> Tested-by: Larry Finger<[email protected]>
> Reviewed-by: Larry Finger<[email protected]>
> Signed-off-by: Ping-Ke Shih <[email protected]>

Patch applied to wireless-next.git, thanks.

47515664ecfb wifi: rtw89: fix potential race condition between napi_init and napi_enable

--
https://patchwork.kernel.org/project/linux-wireless/patch/[email protected]/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches