| Message ID | 20171206200601.GH4283@codeaurora.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 12/06/2017 09:06 PM, Stephen Boyd wrote: > On 12/06, Loys Ollivier wrote: >> When using other platform architectures, in the init of the qcom_scm >> driver, of_node_put is called on /firmware if no qcom dt is found. >> This results in a kernel error: Bad of_node_put() on /firmware. >> >> The call to of_node_put from the qcom_scm init is unnecessary as >> of_find_matching_node is calling it automatically. >> >> Remove this of_node_put(). >> >> Fixes: d0f6fa7ba2d6 ("firmware: qcom: scm: Convert SCM to platform driver") >> Signed-off-by: Loys Ollivier <lollivier@baylibre.com> >> --- > > This still looks wrong. Especially if of_find_matching_node() is > going to look for siblings of the /firmware node for the > compatible string for scm device. Why do we check at all? Can't > we just delete this and let of_platform_populate() take care of > it? BTW, OP-TEE driver seems to have a similar problem. https://lkml.org/lkml/2017/11/29/230 > > ---8<---- > > diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c > index af4c75217ea6..440d8f796faa 100644 > --- a/drivers/firmware/qcom_scm.c > +++ b/drivers/firmware/qcom_scm.c > @@ -626,23 +626,11 @@ static int __init qcom_scm_init(void) > int ret; > > fw_np = of_find_node_by_name(NULL, "firmware"); > - > if (!fw_np) > - return -ENODEV; > - > - np = of_find_matching_node(fw_np, qcom_scm_dt_match); > - > - if (!np) { > - of_node_put(fw_np); > - return -ENODEV; > - } > - > - of_node_put(np); > + return 0; > > ret = of_platform_populate(fw_np, qcom_scm_dt_match, NULL, NULL); > - > of_node_put(fw_np); > - > if (ret) > return ret; > >
On 07/12/2017 09:42, Jerome Forissier wrote: > > > On 12/06/2017 09:06 PM, Stephen Boyd wrote: >> On 12/06, Loys Ollivier wrote: >>> When using other platform architectures, in the init of the qcom_scm >>> driver, of_node_put is called on /firmware if no qcom dt is found. >>> This results in a kernel error: Bad of_node_put() on /firmware. >>> >>> The call to of_node_put from the qcom_scm init is unnecessary as >>> of_find_matching_node is calling it automatically. >>> >>> Remove this of_node_put(). >>> >>> Fixes: d0f6fa7ba2d6 ("firmware: qcom: scm: Convert SCM to platform driver") >>> Signed-off-by: Loys Ollivier <lollivier@baylibre.com> >>> --- >> >> This still looks wrong. Especially if of_find_matching_node() is >> going to look for siblings of the /firmware node for the >> compatible string for scm device. Why do we check at all? Can't >> we just delete this and let of_platform_populate() take care of >> it? BTW, OP-TEE driver seems to have a similar problem. > > https://lkml.org/lkml/2017/11/29/230 > Well, the patch I sent is a fix for a specific bug I am encountering. I tested the patch and it solves my problem. Stephen, your changes looks good but it's a change in the driver's behavior. Maybe it could be another patch ? It looks like OP-TEE had the same idea. >> >> ---8<---- >> >> diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c >> index af4c75217ea6..440d8f796faa 100644 >> --- a/drivers/firmware/qcom_scm.c >> +++ b/drivers/firmware/qcom_scm.c >> @@ -626,23 +626,11 @@ static int __init qcom_scm_init(void) >> int ret; >> >> fw_np = of_find_node_by_name(NULL, "firmware"); >> - >> if (!fw_np) >> - return -ENODEV; >> - >> - np = of_find_matching_node(fw_np, qcom_scm_dt_match); >> - >> - if (!np) { >> - of_node_put(fw_np); >> - return -ENODEV; >> - } >> - >> - of_node_put(np); >> + return 0; >> >> ret = of_platform_populate(fw_np, qcom_scm_dt_match, NULL, NULL); >> - >> of_node_put(fw_np); >> - >> if (ret) >> return ret; >> >>
On 12/07, Loys Ollivier wrote: > > > On 07/12/2017 09:42, Jerome Forissier wrote: > > > > > > On 12/06/2017 09:06 PM, Stephen Boyd wrote: > >> On 12/06, Loys Ollivier wrote: > >>> When using other platform architectures, in the init of the qcom_scm > >>> driver, of_node_put is called on /firmware if no qcom dt is found. > >>> This results in a kernel error: Bad of_node_put() on /firmware. > >>> > >>> The call to of_node_put from the qcom_scm init is unnecessary as > >>> of_find_matching_node is calling it automatically. > >>> > >>> Remove this of_node_put(). > >>> > >>> Fixes: d0f6fa7ba2d6 ("firmware: qcom: scm: Convert SCM to platform driver") > >>> Signed-off-by: Loys Ollivier <lollivier@baylibre.com> > >>> --- > >> > >> This still looks wrong. Especially if of_find_matching_node() is > >> going to look for siblings of the /firmware node for the > >> compatible string for scm device. Why do we check at all? Can't > >> we just delete this and let of_platform_populate() take care of > >> it? BTW, OP-TEE driver seems to have a similar problem. > > > > https://lkml.org/lkml/2017/11/29/230 > > > Well, the patch I sent is a fix for a specific bug I am encountering. > I tested the patch and it solves my problem. Stephen, your changes looks > good but it's a change in the driver's behavior. Maybe it could be > another patch ? Sure. But there's another of_node_put(fw_np) in this function, so why isn't that also removed? Assuming of_find_matching_node() is calling of_node_put() on what's passed in, then the node is going to get put twice in the "working" case. Andy?
On 08/12/2017 18:04, Stephen Boyd wrote: > On 12/07, Loys Ollivier wrote: >> >> >> On 07/12/2017 09:42, Jerome Forissier wrote: >>> >>> >>> On 12/06/2017 09:06 PM, Stephen Boyd wrote: >>>> On 12/06, Loys Ollivier wrote: >>>>> When using other platform architectures, in the init of the qcom_scm >>>>> driver, of_node_put is called on /firmware if no qcom dt is found. >>>>> This results in a kernel error: Bad of_node_put() on /firmware. >>>>> >>>>> The call to of_node_put from the qcom_scm init is unnecessary as >>>>> of_find_matching_node is calling it automatically. >>>>> >>>>> Remove this of_node_put(). >>>>> >>>>> Fixes: d0f6fa7ba2d6 ("firmware: qcom: scm: Convert SCM to platform driver") >>>>> Signed-off-by: Loys Ollivier <lollivier@baylibre.com> >>>>> --- >>>> >>>> This still looks wrong. Especially if of_find_matching_node() is >>>> going to look for siblings of the /firmware node for the >>>> compatible string for scm device. Why do we check at all? Can't >>>> we just delete this and let of_platform_populate() take care of >>>> it? BTW, OP-TEE driver seems to have a similar problem. >>> >>> https://lkml.org/lkml/2017/11/29/230 >>> >> Well, the patch I sent is a fix for a specific bug I am encountering. >> I tested the patch and it solves my problem. Stephen, your changes looks >> good but it's a change in the driver's behavior. Maybe it could be >> another patch ? > > Sure. But there's another of_node_put(fw_np) in this function, so > why isn't that also removed? Assuming of_find_matching_node() is > calling of_node_put() on what's passed in, then the node is going > to get put twice in the "working" case. > > Andy? > Agreed, I had a look and this second call to of_node_put(fw_np) seem to be unnecessary as well. Unfortunately I can't test your suggestion as I am using another platform arch. I am just testing that this driver does not break my arch. I can submit a v3 removing this of_node_put as well if you want.
diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c index af4c75217ea6..440d8f796faa 100644 --- a/drivers/firmware/qcom_scm.c +++ b/drivers/firmware/qcom_scm.c @@ -626,23 +626,11 @@ static int __init qcom_scm_init(void) int ret; fw_np = of_find_node_by_name(NULL, "firmware"); - if (!fw_np) - return -ENODEV; - - np = of_find_matching_node(fw_np, qcom_scm_dt_match); - - if (!np) { - of_node_put(fw_np); - return -ENODEV; - } - - of_node_put(np); + return 0; ret = of_platform_populate(fw_np, qcom_scm_dt_match, NULL, NULL); - of_node_put(fw_np); - if (ret) return ret;