| Message ID | 20170210215525.139701-1-briannorris@chromium.org (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 2447e2cad75239ae407c0f98acf12511354208c5 |
| Delegated to: | Kalle Valo |
| Headers | show |
Brian Norris <briannorris@chromium.org> writes: > If we don't have an out-of-band wakeup IRQ configured through DT (as > most platforms don't), then we fall out of this function with > 'irq_wakeup == 0'. Other code (e.g., mwifiex_disable_wake() and > mwifiex_enable_wake()) treats 'irq_wakeup >= 0' as a valid IRQ, and so > we end up calling {enable,disable}_irq() on IRQ 0. > > That seems bad, so let's not do that. > > Same problem as fixed in this patch: > > https://patchwork.kernel.org/patch/9531693/ > [PATCH v2 2/3] btmrvl: set irq_bt to -1 when failed to parse it > > with the difference that: > (a) this one is actually a regression and > (b) this affects both device tree and non-device-tree systems > > While fixing the regression, also drop the verbosity on the parse > failure, so we don't see this when a DT node is present but doesn't have > an interrupt property (this is perfectly legal): > > [ 21.999000] mwifiex_pcie 0000:01:00.0: fail to parse irq_wakeup from device tree > > Fixes: 853402a00823 ("mwifiex: Enable WoWLAN for both sdio and pcie") > Signed-off-by: Brian Norris <briannorris@chromium.org> > --- > This is a 4.10 regression. Not sure if it's critical to push quickly, or if it > should just be tagged -stable and pushed in 4.11. It's really late for 4.10, so I would rather push it to 4.11.
On Fri, Feb 10, 2017 at 1:55 PM, Brian Norris <briannorris@chromium.org> wrote: > If we don't have an out-of-band wakeup IRQ configured through DT (as > most platforms don't), then we fall out of this function with > 'irq_wakeup == 0'. Other code (e.g., mwifiex_disable_wake() and > mwifiex_enable_wake()) treats 'irq_wakeup >= 0' as a valid IRQ, and so > we end up calling {enable,disable}_irq() on IRQ 0. > > That seems bad, so let's not do that. > > Same problem as fixed in this patch: > > https://patchwork.kernel.org/patch/9531693/ > [PATCH v2 2/3] btmrvl: set irq_bt to -1 when failed to parse it > > with the difference that: > (a) this one is actually a regression and > (b) this affects both device tree and non-device-tree systems > > While fixing the regression, also drop the verbosity on the parse > failure, so we don't see this when a DT node is present but doesn't have > an interrupt property (this is perfectly legal): > > [ 21.999000] mwifiex_pcie 0000:01:00.0: fail to parse irq_wakeup from device tree > > Fixes: 853402a00823 ("mwifiex: Enable WoWLAN for both sdio and pcie") > Signed-off-by: Brian Norris <briannorris@chromium.org> Acked-by: Rajat Jain <rajatja@google.com> > --- > This is a 4.10 regression. Not sure if it's critical to push quickly, or if it > should just be tagged -stable and pushed in 4.11. > > drivers/net/wireless/marvell/mwifiex/main.c | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/net/wireless/marvell/mwifiex/main.c b/drivers/net/wireless/marvell/mwifiex/main.c > index e5c3a8aa3929..c246d8efb6fc 100644 > --- a/drivers/net/wireless/marvell/mwifiex/main.c > +++ b/drivers/net/wireless/marvell/mwifiex/main.c > @@ -1569,13 +1569,13 @@ static void mwifiex_probe_of(struct mwifiex_adapter *adapter) > struct device *dev = adapter->dev; > > if (!dev->of_node) > - return; > + goto err_exit; > > adapter->dt_node = dev->of_node; > adapter->irq_wakeup = irq_of_parse_and_map(adapter->dt_node, 0); > if (!adapter->irq_wakeup) { > - dev_info(dev, "fail to parse irq_wakeup from device tree\n"); > - return; > + dev_dbg(dev, "fail to parse irq_wakeup from device tree\n"); > + goto err_exit; > } > > ret = devm_request_irq(dev, adapter->irq_wakeup, > @@ -1595,7 +1595,7 @@ static void mwifiex_probe_of(struct mwifiex_adapter *adapter) > return; > > err_exit: > - adapter->irq_wakeup = 0; > + adapter->irq_wakeup = -1; > } > > /* > -- > 2.11.0.483.g087da7b7c-goog >
Brian Norris <briannorris@chromium.org> wrote: > If we don't have an out-of-band wakeup IRQ configured through DT (as > most platforms don't), then we fall out of this function with > 'irq_wakeup == 0'. Other code (e.g., mwifiex_disable_wake() and > mwifiex_enable_wake()) treats 'irq_wakeup >= 0' as a valid IRQ, and so > we end up calling {enable,disable}_irq() on IRQ 0. > > That seems bad, so let's not do that. > > Same problem as fixed in this patch: > > https://patchwork.kernel.org/patch/9531693/ > [PATCH v2 2/3] btmrvl: set irq_bt to -1 when failed to parse it > > with the difference that: > (a) this one is actually a regression and > (b) this affects both device tree and non-device-tree systems > > While fixing the regression, also drop the verbosity on the parse > failure, so we don't see this when a DT node is present but doesn't have > an interrupt property (this is perfectly legal): > > [ 21.999000] mwifiex_pcie 0000:01:00.0: fail to parse irq_wakeup from device tree > > Fixes: 853402a00823 ("mwifiex: Enable WoWLAN for both sdio and pcie") > Signed-off-by: Brian Norris <briannorris@chromium.org> > Acked-by: Rajat Jain <rajatja@google.com> Patch applied to wireless-drivers-next.git, thanks. 2447e2cad752 mwifiex: don't enable/disable IRQ 0 during suspend/resume
diff --git a/drivers/net/wireless/marvell/mwifiex/main.c b/drivers/net/wireless/marvell/mwifiex/main.c index e5c3a8aa3929..c246d8efb6fc 100644 --- a/drivers/net/wireless/marvell/mwifiex/main.c +++ b/drivers/net/wireless/marvell/mwifiex/main.c @@ -1569,13 +1569,13 @@ static void mwifiex_probe_of(struct mwifiex_adapter *adapter) struct device *dev = adapter->dev; if (!dev->of_node) - return; + goto err_exit; adapter->dt_node = dev->of_node; adapter->irq_wakeup = irq_of_parse_and_map(adapter->dt_node, 0); if (!adapter->irq_wakeup) { - dev_info(dev, "fail to parse irq_wakeup from device tree\n"); - return; + dev_dbg(dev, "fail to parse irq_wakeup from device tree\n"); + goto err_exit; } ret = devm_request_irq(dev, adapter->irq_wakeup, @@ -1595,7 +1595,7 @@ static void mwifiex_probe_of(struct mwifiex_adapter *adapter) return; err_exit: - adapter->irq_wakeup = 0; + adapter->irq_wakeup = -1; } /*
If we don't have an out-of-band wakeup IRQ configured through DT (as most platforms don't), then we fall out of this function with 'irq_wakeup == 0'. Other code (e.g., mwifiex_disable_wake() and mwifiex_enable_wake()) treats 'irq_wakeup >= 0' as a valid IRQ, and so we end up calling {enable,disable}_irq() on IRQ 0. That seems bad, so let's not do that. Same problem as fixed in this patch: https://patchwork.kernel.org/patch/9531693/ [PATCH v2 2/3] btmrvl: set irq_bt to -1 when failed to parse it with the difference that: (a) this one is actually a regression and (b) this affects both device tree and non-device-tree systems While fixing the regression, also drop the verbosity on the parse failure, so we don't see this when a DT node is present but doesn't have an interrupt property (this is perfectly legal): [ 21.999000] mwifiex_pcie 0000:01:00.0: fail to parse irq_wakeup from device tree Fixes: 853402a00823 ("mwifiex: Enable WoWLAN for both sdio and pcie") Signed-off-by: Brian Norris <briannorris@chromium.org> --- This is a 4.10 regression. Not sure if it's critical to push quickly, or if it should just be tagged -stable and pushed in 4.11. drivers/net/wireless/marvell/mwifiex/main.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)