| Message ID | 20191121213935.2cbgh3qmd4hv4v5a@kili.mountain (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | b43e36d75e8727f78892652a25967a1ffa03d1d1 |
| Delegated to: | Kalle Valo |
| Headers | show |
| Series | mt76: Off by one in mt76_calc_rx_airtime() | expand |
Dan Carpenter <dan.carpenter@oracle.com> writes: > The sband->bitrates[] array has "sband->n_bitrates" elements so this > check needs to be >= instead of > or we could read beyond the end of the > array. > > These values come from when we call mt76_register_device(): > > ret = mt76_register_device(&dev->mt76, true, mt7603_rates, > ARRAY_SIZE(mt7603_rates)); > > Here sband->bitrates[] is mt7603_rates[] and ->n_bitrates is the > ARRAY_SIZE() > > Fixes: 5ce09c1a7907 ("mt76: track rx airtime for airtime fairness and survey") > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Should I queue this to v5.5?
On 11/26/19 5:49 AM, Dan Carpenter wrote: > The sband->bitrates[] array has "sband->n_bitrates" elements so this > check needs to be >= instead of > or we could read beyond the end of the > array. > > These values come from when we call mt76_register_device(): > > ret = mt76_register_device(&dev->mt76, true, mt7603_rates, > ARRAY_SIZE(mt7603_rates)); > > Here sband->bitrates[] is mt7603_rates[] and ->n_bitrates is the > ARRAY_SIZE() > > Fixes: 5ce09c1a7907 ("mt76: track rx airtime for airtime fairness and survey") > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> > --- > drivers/net/wireless/mediatek/mt76/airtime.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/net/wireless/mediatek/mt76/airtime.c b/drivers/net/wireless/mediatek/mt76/airtime.c > index 55116f395f9a..a4a785467748 100644 > --- a/drivers/net/wireless/mediatek/mt76/airtime.c > +++ b/drivers/net/wireless/mediatek/mt76/airtime.c > @@ -242,7 +242,7 @@ u32 mt76_calc_rx_airtime(struct mt76_dev *dev, struct mt76_rx_status *status, > return 0; > > sband = dev->hw->wiphy->bands[status->band]; > - if (!sband || status->rate_idx > sband->n_bitrates) > + if (!sband || status->rate_idx >= sband->n_bitrates) > return 0; > > rate = &sband->bitrates[status->rate_idx]; This should fix the division by zero error I sometimes got yesterday. [ 63.115806] divide error: 0000 [#1] PREEMPT SMP NOPTI [ 63.121054] CPU: 0 PID: 225 Comm: kworker/u8:2 Tainted: G W 5.4.0-rc7-1-01110-g19b7e21c55c8 #39 [ 63.131331] Hardware name: PC Engines apu2/apu2, BIOS v4.10.0.3 11/07/2019 [ 63.138399] Workqueue: mt76u mt76u_tx_status_data [mt76_usb] [ 63.144258] RIP: 0010:mt76_calc_rx_airtime+0x12b/0x150 [mt76] [ 63.150281] Code: 8d 34 76 48 8d 34 b1 0f b6 4e 07 66 85 c9 74 25 66 83 f9 01 75 1c b9 24 00 00 00 89 d0 0f b7 76 04 c1 e0 05 8d 04 d0 01 c0 99 <f7> fe 01 c8 c3 31 c0 c3 0f 0b c3 44 89 c8 83 e0 01 3c 01 19 c9 83 [ 63.169699] RSP: 0018:ffffbd694032fcf0 EFLAGS: 00010216 [ 63.175176] RAX: 0000000000002850 RBX: 0000000000000001 RCX: 00000000000000ca [ 63.182535] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffffa20068253538 [ 63.189983] RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000000000000 [ 63.197381] R10: 0000000000000000 R11: ffffbd694032fcb0 R12: ffffbd694032fdb0 [ 63.204782] R13: 0000000000000000 R14: ffffbd694032fd00 R15: ffffa20068251e40 [ 63.212228] FS: 0000000000000000(0000) GS:ffffa2006aa00000(0000) knlGS:0000000000000000 [ 63.220614] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 63.226500] CR2: 00007fe65ba85cb0 CR3: 00000001293b6000 CR4: 00000000000406f0 [ 63.233841] Call Trace: [ 63.236410] mt76_calc_tx_airtime+0xf4/0x190 [mt76] [ 63.241464] mt76x02_send_tx_status+0x1cd/0x3f0 [mt76x02_lib] [ 63.247430] mt76x02_tx_status_data+0x54/0x80 [mt76x02_lib] [ 63.253186] mt76u_tx_status_data+0x63/0xc0 [mt76_usb] [ 63.258451] process_one_work+0x1e2/0x3b0 [ 63.262533] worker_thread+0x4a/0x3d0 [ 63.266306] kthread+0xfb/0x130 [ 63.269550] ? process_one_work+0x3b0/0x3b0 [ 63.273893] ? kthread_park+0x90/0x90 [ 63.277677] ret_from_fork+0x22/0x40 [ 63.281411] Modules linked in: ccm bridge stp llc mt76x2u mt76x2_common mt76x02_usb mt76_usb mt76x02_lib mt76 nft_masq nft_chain_nat nf_nat nft_reject_inet nf_reject_ipv4 nf_reject_ipv6 nft_reject nft_ct nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 libcrc32c nf_tables_set nf_tables nfnetlink ath10k_pci ath10k_core amd64_edac_mod edac_mce_amd ath kvm_amd mac80211 kvm irqbypass crct10dif_pclmul crc32_pclmul ghash_clmulni_intel aesni_intel pcengines_apuv2 cfg80211 gpio_keys_polled crypto_simd input_polldev gpio_amd_fch cryptd igb glue_helper pcspkr fam15h_power sp5100_tco k10temp i2c_piix4 rfkill i2c_algo_bit ccp libarc4 dca rng_core uio_pdrv_genirq evdev leds_gpio uio mac_hid coreboot_table acpi_cpufreq pinctrl_amd sr_mod cdrom ip_tables x_tables ext4 crc32c_generic crc16 mbcache jbd2 sd_mod usb_storage ahci libahci libata xhci_pci sdhci_pci xhci_hcd scsi_mod cqhci sdhci ehci_pci crc32c_intel ehci_hcd mmc_core gpio_keys [ 63.365937] ---[ end trace f13e9cdc5f55db9e ]--- [ 63.370802] RIP: 0010:mt76_calc_rx_airtime+0x12b/0x150 [mt76] [ 63.376807] Code: 8d 34 76 48 8d 34 b1 0f b6 4e 07 66 85 c9 74 25 66 83 f9 01 75 1c b9 24 00 00 00 89 d0 0f b7 76 04 c1 e0 05 8d 04 d0 01 c0 99 <f7> fe 01 c8 c3 31 c0 c3 0f 0b c3 44 89 c8 83 e0 01 3c 01 19 c9 83 [ 63.396220] RSP: 0018:ffffbd694032fcf0 EFLAGS: 00010216 [ 63.401660] RAX: 0000000000002850 RBX: 0000000000000001 RCX: 00000000000000ca [ 63.409145] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffffa20068253538 [ 63.416505] RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000000000000 [ 63.423988] R10: 0000000000000000 R11: ffffbd694032fcb0 R12: ffffbd694032fdb0 [ 63.431425] R13: 0000000000000000 R14: ffffbd694032fd00 R15: ffffa20068251e40 [ 63.438793] FS: 0000000000000000(0000) GS:ffffa2006aa00000(0000) knlGS:0000000000000000 [ 63.447141] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 63.453216] CR2: 00007fe65ba85cb0 CR3: 00000001293b6000 CR4: 00000000000406f0
Dan Carpenter <dan.carpenter@oracle.com> writes: > The sband->bitrates[] array has "sband->n_bitrates" elements so this > check needs to be >= instead of > or we could read beyond the end of the > array. > > These values come from when we call mt76_register_device(): > > ret = mt76_register_device(&dev->mt76, true, mt7603_rates, > ARRAY_SIZE(mt7603_rates)); > > Here sband->bitrates[] is mt7603_rates[] and ->n_bitrates is the > ARRAY_SIZE() > > Fixes: 5ce09c1a7907 ("mt76: track rx airtime for airtime fairness and survey") > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> > --- > drivers/net/wireless/mediatek/mt76/airtime.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/net/wireless/mediatek/mt76/airtime.c b/drivers/net/wireless/mediatek/mt76/airtime.c > index 55116f395f9a..a4a785467748 100644 > --- a/drivers/net/wireless/mediatek/mt76/airtime.c > +++ b/drivers/net/wireless/mediatek/mt76/airtime.c > @@ -242,7 +242,7 @@ u32 mt76_calc_rx_airtime(struct mt76_dev *dev, struct mt76_rx_status *status, > return 0; > > sband = dev->hw->wiphy->bands[status->band]; > - if (!sband || status->rate_idx > sband->n_bitrates) > + if (!sband || status->rate_idx >= sband->n_bitrates) > return 0; > > rate = &sband->bitrates[status->rate_idx]; This code has recently been ported to mac80211 (net/mac80211/airtime.c). It seems that the bug is also present there; care to send a patch for that as well? :) -Toke
On Tue, Nov 26, 2019 at 09:04:15AM +0100, Toke Høiland-Jørgensen wrote: > Dan Carpenter <dan.carpenter@oracle.com> writes: > > > The sband->bitrates[] array has "sband->n_bitrates" elements so this > > check needs to be >= instead of > or we could read beyond the end of the > > array. > > > > These values come from when we call mt76_register_device(): > > > > ret = mt76_register_device(&dev->mt76, true, mt7603_rates, > > ARRAY_SIZE(mt7603_rates)); > > > > Here sband->bitrates[] is mt7603_rates[] and ->n_bitrates is the > > ARRAY_SIZE() > > > > Fixes: 5ce09c1a7907 ("mt76: track rx airtime for airtime fairness and survey") > > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> > > --- > > drivers/net/wireless/mediatek/mt76/airtime.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/drivers/net/wireless/mediatek/mt76/airtime.c b/drivers/net/wireless/mediatek/mt76/airtime.c > > index 55116f395f9a..a4a785467748 100644 > > --- a/drivers/net/wireless/mediatek/mt76/airtime.c > > +++ b/drivers/net/wireless/mediatek/mt76/airtime.c > > @@ -242,7 +242,7 @@ u32 mt76_calc_rx_airtime(struct mt76_dev *dev, struct mt76_rx_status *status, > > return 0; > > > > sband = dev->hw->wiphy->bands[status->band]; > > - if (!sband || status->rate_idx > sband->n_bitrates) > > + if (!sband || status->rate_idx >= sband->n_bitrates) > > return 0; > > > > rate = &sband->bitrates[status->rate_idx]; > > This code has recently been ported to mac80211 (net/mac80211/airtime.c). > It seems that the bug is also present there; care to send a patch for > that as well? :) Oh. Thanks for pointing that out. I actually saw the static checker warning for that and ignored it thinking that it was the same code. :P I will send a fix for it. regards, dan carpenter
Dan Carpenter <dan.carpenter@oracle.com> writes: > On Tue, Nov 26, 2019 at 09:04:15AM +0100, Toke Høiland-Jørgensen wrote: >> Dan Carpenter <dan.carpenter@oracle.com> writes: >> >> > The sband->bitrates[] array has "sband->n_bitrates" elements so this >> > check needs to be >= instead of > or we could read beyond the end of the >> > array. >> > >> > These values come from when we call mt76_register_device(): >> > >> > ret = mt76_register_device(&dev->mt76, true, mt7603_rates, >> > ARRAY_SIZE(mt7603_rates)); >> > >> > Here sband->bitrates[] is mt7603_rates[] and ->n_bitrates is the >> > ARRAY_SIZE() >> > >> > Fixes: 5ce09c1a7907 ("mt76: track rx airtime for airtime fairness and survey") >> > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> >> > --- >> > drivers/net/wireless/mediatek/mt76/airtime.c | 2 +- >> > 1 file changed, 1 insertion(+), 1 deletion(-) >> > >> > diff --git a/drivers/net/wireless/mediatek/mt76/airtime.c b/drivers/net/wireless/mediatek/mt76/airtime.c >> > index 55116f395f9a..a4a785467748 100644 >> > --- a/drivers/net/wireless/mediatek/mt76/airtime.c >> > +++ b/drivers/net/wireless/mediatek/mt76/airtime.c >> > @@ -242,7 +242,7 @@ u32 mt76_calc_rx_airtime(struct mt76_dev *dev, struct mt76_rx_status *status, >> > return 0; >> > >> > sband = dev->hw->wiphy->bands[status->band]; >> > - if (!sband || status->rate_idx > sband->n_bitrates) >> > + if (!sband || status->rate_idx >= sband->n_bitrates) >> > return 0; >> > >> > rate = &sband->bitrates[status->rate_idx]; >> >> This code has recently been ported to mac80211 (net/mac80211/airtime.c). >> It seems that the bug is also present there; care to send a patch for >> that as well? :) > > Oh. Thanks for pointing that out. I actually saw the static checker > warning for that and ignored it thinking that it was the same code. > :P Well, it's copy-pasted from the same code ;) The plan is to get rid of the version inside mt76; was waiting for the trees to converge, though, so I guess after the merge window? > I will send a fix for it. Great, thanks! -Toke
Toke Høiland-Jørgensen <toke@redhat.com> writes: > Dan Carpenter <dan.carpenter@oracle.com> writes: > >> On Tue, Nov 26, 2019 at 09:04:15AM +0100, Toke Høiland-Jørgensen wrote: >>> Dan Carpenter <dan.carpenter@oracle.com> writes: >>> >>> > @@ -242,7 +242,7 @@ u32 mt76_calc_rx_airtime(struct mt76_dev *dev, struct mt76_rx_status *status, >>> > return 0; >>> > >>> > sband = dev->hw->wiphy->bands[status->band]; >>> > - if (!sband || status->rate_idx > sband->n_bitrates) >>> > + if (!sband || status->rate_idx >= sband->n_bitrates) >>> > return 0; >>> > >>> > rate = &sband->bitrates[status->rate_idx]; >>> >>> This code has recently been ported to mac80211 (net/mac80211/airtime.c). >>> It seems that the bug is also present there; care to send a patch for >>> that as well? :) >> >> Oh. Thanks for pointing that out. I actually saw the static checker >> warning for that and ignored it thinking that it was the same code. >> :P > > Well, it's copy-pasted from the same code ;) > > The plan is to get rid of the version inside mt76; was waiting for the > trees to converge, though, so I guess after the merge window? Yup, as I'm not taking anything to w-d-next until -rc1 is released.
Dan Carpenter <dan.carpenter@oracle.com> wrote: > The sband->bitrates[] array has "sband->n_bitrates" elements so this > check needs to be >= instead of > or we could read beyond the end of the > array. > > These values come from when we call mt76_register_device(): > > ret = mt76_register_device(&dev->mt76, true, mt7603_rates, > ARRAY_SIZE(mt7603_rates)); > > Here sband->bitrates[] is mt7603_rates[] and ->n_bitrates is the > ARRAY_SIZE() > > Fixes: 5ce09c1a7907 ("mt76: track rx airtime for airtime fairness and survey") > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Patch applied to wireless-drivers.git, thanks. b43e36d75e87 mt76: Off by one in mt76_calc_rx_airtime()
diff --git a/drivers/net/wireless/mediatek/mt76/airtime.c b/drivers/net/wireless/mediatek/mt76/airtime.c index 55116f395f9a..a4a785467748 100644 --- a/drivers/net/wireless/mediatek/mt76/airtime.c +++ b/drivers/net/wireless/mediatek/mt76/airtime.c @@ -242,7 +242,7 @@ u32 mt76_calc_rx_airtime(struct mt76_dev *dev, struct mt76_rx_status *status, return 0; sband = dev->hw->wiphy->bands[status->band]; - if (!sband || status->rate_idx > sband->n_bitrates) + if (!sband || status->rate_idx >= sband->n_bitrates) return 0; rate = &sband->bitrates[status->rate_idx];
The sband->bitrates[] array has "sband->n_bitrates" elements so this check needs to be >= instead of > or we could read beyond the end of the array. These values come from when we call mt76_register_device(): ret = mt76_register_device(&dev->mt76, true, mt7603_rates, ARRAY_SIZE(mt7603_rates)); Here sband->bitrates[] is mt7603_rates[] and ->n_bitrates is the ARRAY_SIZE() Fixes: 5ce09c1a7907 ("mt76: track rx airtime for airtime fairness and survey") Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> --- drivers/net/wireless/mediatek/mt76/airtime.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)