Failure to register SGI for firmware event notification is non-fatal error
when feature is not supported by other modules such as Xen and TF-A. Add
_info level log message for such special case.
Also add XST_PM_INVALID_VERSION error code and map it to -EOPNOSUPP Linux
kernel error code. If feature is not supported or EEMI API version is
mismatch, firmware can return XST_PM_INVALID_VERSION = 4 or
XST_PM_NO_FEATURE = 19 error code.
Co-developed-by: Tanmay Shah <tanmay.shah@amd.com>
Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>
Signed-off-by: Jay Buddhabhatti <jay.buddhabhatti@amd.com>
---
drivers/firmware/xilinx/zynqmp.c | 4 +++-
drivers/soc/xilinx/xlnx_event_manager.c | 7 ++++++-
include/linux/firmware/xlnx-zynqmp.h | 1 +
3 files changed, 10 insertions(+), 2 deletions(-)
diff --git a/drivers/firmware/xilinx/zynqmp.c b/drivers/firmware/xilinx/zynqmp.c
index 00ea2508f692..de09d77cf1e4 100644
--- a/drivers/firmware/xilinx/zynqmp.c
+++ b/drivers/firmware/xilinx/zynqmp.c
@@ -92,6 +92,8 @@ static int zynqmp_pm_ret_code(u32 ret_status)
return 0;
case XST_PM_NO_FEATURE:
return -ENOTSUPP;
+ case XST_PM_INVALID_VERSION:
+ return -EOPNOTSUPP;
case XST_PM_NO_ACCESS:
return -EACCES;
case XST_PM_ABORT_SUSPEND:
@@ -379,7 +381,7 @@ int zynqmp_pm_register_sgi(u32 sgi_num, u32 reset)
ret = zynqmp_pm_invoke_fn(TF_A_PM_REGISTER_SGI, sgi_num, reset, 0, 0,
0, NULL);
- if (!ret)
+ if (ret != -EOPNOTSUPP && !ret)
return ret;
/* try old implementation as fallback strategy if above fails */
diff --git a/drivers/soc/xilinx/xlnx_event_manager.c b/drivers/soc/xilinx/xlnx_event_manager.c
index 38cfc161a713..bfa5cf4a0cab 100644
--- a/drivers/soc/xilinx/xlnx_event_manager.c
+++ b/drivers/soc/xilinx/xlnx_event_manager.c
@@ -653,7 +653,12 @@ static int xlnx_event_manager_probe(struct platform_device *pdev)
ret = zynqmp_pm_register_sgi(sgi_num, 0);
if (ret) {
- dev_err(&pdev->dev, "SGI %d Registration over TF-A failed with %d\n", sgi_num, ret);
+ if (ret == -EOPNOTSUPP)
+ dev_info(&pdev->dev, "PM firmware event notification not supported\n");
+ else
+ dev_err(&pdev->dev, "SGI %d registration failed, err %d\n",
+ sgi_num, ret);
+
xlnx_event_cleanup_sgi(pdev);
return ret;
}
diff --git a/include/linux/firmware/xlnx-zynqmp.h b/include/linux/firmware/xlnx-zynqmp.h
index 86fb73269d95..4afd5d63177e 100644
--- a/include/linux/firmware/xlnx-zynqmp.h
+++ b/include/linux/firmware/xlnx-zynqmp.h
@@ -160,6 +160,7 @@ enum pm_api_id {
/* PMU-FW return status codes */
enum pm_ret_status {
XST_PM_SUCCESS = 0,
+ XST_PM_INVALID_VERSION = 4,
XST_PM_NO_FEATURE = 19,
XST_PM_INTERNAL = 2000,
XST_PM_CONFLICT = 2001,
--
2.25.1
On Wed, Oct 04, 2023 at 02:41:14AM -0700, Jay Buddhabhatti wrote:
> Failure to register SGI for firmware event notification is non-fatal error
> when feature is not supported by other modules such as Xen and TF-A. Add
> _info level log message for such special case.
>
> Also add XST_PM_INVALID_VERSION error code and map it to -EOPNOSUPP Linux
> kernel error code. If feature is not supported or EEMI API version is
> mismatch, firmware can return XST_PM_INVALID_VERSION = 4 or
> XST_PM_NO_FEATURE = 19 error code.
>
> Co-developed-by: Tanmay Shah <tanmay.shah@amd.com>
> Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>
> Signed-off-by: Jay Buddhabhatti <jay.buddhabhatti@amd.com>
> ---
> drivers/firmware/xilinx/zynqmp.c | 4 +++-
> drivers/soc/xilinx/xlnx_event_manager.c | 7 ++++++-
> include/linux/firmware/xlnx-zynqmp.h | 1 +
> 3 files changed, 10 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/firmware/xilinx/zynqmp.c b/drivers/firmware/xilinx/zynqmp.c
> index 00ea2508f692..de09d77cf1e4 100644
> --- a/drivers/firmware/xilinx/zynqmp.c
> +++ b/drivers/firmware/xilinx/zynqmp.c
> @@ -92,6 +92,8 @@ static int zynqmp_pm_ret_code(u32 ret_status)
> return 0;
> case XST_PM_NO_FEATURE:
> return -ENOTSUPP;
> + case XST_PM_INVALID_VERSION:
> + return -EOPNOTSUPP;
> case XST_PM_NO_ACCESS:
> return -EACCES;
> case XST_PM_ABORT_SUSPEND:
> @@ -379,7 +381,7 @@ int zynqmp_pm_register_sgi(u32 sgi_num, u32 reset)
>
> ret = zynqmp_pm_invoke_fn(TF_A_PM_REGISTER_SGI, sgi_num, reset, 0, 0,
> 0, NULL);
> - if (!ret)
> + if (ret != -EOPNOTSUPP && !ret)
> return ret;
>
> /* try old implementation as fallback strategy if above fails */
> diff --git a/drivers/soc/xilinx/xlnx_event_manager.c b/drivers/soc/xilinx/xlnx_event_manager.c
> index 38cfc161a713..bfa5cf4a0cab 100644
> --- a/drivers/soc/xilinx/xlnx_event_manager.c
> +++ b/drivers/soc/xilinx/xlnx_event_manager.c
> @@ -653,7 +653,12 @@ static int xlnx_event_manager_probe(struct platform_device *pdev)
>
> ret = zynqmp_pm_register_sgi(sgi_num, 0);
> if (ret) {
> - dev_err(&pdev->dev, "SGI %d Registration over TF-A failed with %d\n", sgi_num, ret);
> + if (ret == -EOPNOTSUPP)
> + dev_info(&pdev->dev, "PM firmware event notification not supported\n");
Why is this "info"? Why not error?
And what is userspace supposed to do with this information if it sees
it?
thanks,
greg k-h
Hi Greg,
> -----Original Message-----
> From: Greg KH <gregkh@linuxfoundation.org>
> Sent: Wednesday, October 4, 2023 5:45 PM
> To: Buddhabhatti, Jay <jay.buddhabhatti@amd.com>
> Cc: Simek, Michal <michal.simek@amd.com>; Shah, Tanmay
> <tanmay.shah@amd.com>; Manne, Nava kishore
> <nava.kishore.manne@amd.com>; Levinsky, Ben <ben.levinsky@amd.com>;
> robh@kernel.org; dhaval.r.shah@amd.com; marex@denx.de; Shaikh, Izhar
> Ameer <izhar.ameer.shaikh@amd.com>; arnd@arndb.de;
> ruanjinjie@huawei.com; mathieu.poirier@linaro.org; linux-
> kernel@vger.kernel.org; linux-arm-kernel@lists.infradead.org
> Subject: Re: [PATCH v2 4/6] drivers: soc: xilinx: Fix error message on SGI
> registration failure
>
> On Wed, Oct 04, 2023 at 02:41:14AM -0700, Jay Buddhabhatti wrote:
> > Failure to register SGI for firmware event notification is non-fatal
> > error when feature is not supported by other modules such as Xen and
> > TF-A. Add _info level log message for such special case.
> >
> > Also add XST_PM_INVALID_VERSION error code and map it to -EOPNOSUPP
> > Linux kernel error code. If feature is not supported or EEMI API
> > version is mismatch, firmware can return XST_PM_INVALID_VERSION = 4 or
> > XST_PM_NO_FEATURE = 19 error code.
> >
> > Co-developed-by: Tanmay Shah <tanmay.shah@amd.com>
> > Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>
> > Signed-off-by: Jay Buddhabhatti <jay.buddhabhatti@amd.com>
> > ---
> > drivers/firmware/xilinx/zynqmp.c | 4 +++-
> > drivers/soc/xilinx/xlnx_event_manager.c | 7 ++++++-
> > include/linux/firmware/xlnx-zynqmp.h | 1 +
> > 3 files changed, 10 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/firmware/xilinx/zynqmp.c
> > b/drivers/firmware/xilinx/zynqmp.c
> > index 00ea2508f692..de09d77cf1e4 100644
> > --- a/drivers/firmware/xilinx/zynqmp.c
> > +++ b/drivers/firmware/xilinx/zynqmp.c
> > @@ -92,6 +92,8 @@ static int zynqmp_pm_ret_code(u32 ret_status)
> > return 0;
> > case XST_PM_NO_FEATURE:
> > return -ENOTSUPP;
> > + case XST_PM_INVALID_VERSION:
> > + return -EOPNOTSUPP;
> > case XST_PM_NO_ACCESS:
> > return -EACCES;
> > case XST_PM_ABORT_SUSPEND:
> > @@ -379,7 +381,7 @@ int zynqmp_pm_register_sgi(u32 sgi_num, u32 reset)
> >
> > ret = zynqmp_pm_invoke_fn(TF_A_PM_REGISTER_SGI, sgi_num, reset,
> 0, 0,
> > 0, NULL);
> > - if (!ret)
> > + if (ret != -EOPNOTSUPP && !ret)
> > return ret;
> >
> > /* try old implementation as fallback strategy if above fails */
> > diff --git a/drivers/soc/xilinx/xlnx_event_manager.c
> > b/drivers/soc/xilinx/xlnx_event_manager.c
> > index 38cfc161a713..bfa5cf4a0cab 100644
> > --- a/drivers/soc/xilinx/xlnx_event_manager.c
> > +++ b/drivers/soc/xilinx/xlnx_event_manager.c
> > @@ -653,7 +653,12 @@ static int xlnx_event_manager_probe(struct
> > platform_device *pdev)
> >
> > ret = zynqmp_pm_register_sgi(sgi_num, 0);
> > if (ret) {
> > - dev_err(&pdev->dev, "SGI %d Registration over TF-A failed with
> %d\n", sgi_num, ret);
> > + if (ret == -EOPNOTSUPP)
> > + dev_info(&pdev->dev, "PM firmware event notification
> not
> > +supported\n");
>
> Why is this "info"? Why not error?
[Jay] Because it is non-fatal error which will not cause kernel crash.
>
> And what is userspace supposed to do with this information if it sees it?
[Jay] This message indicates that SGI registration is not supported by other modules like TF-A or Xen. So user need to use TF-A with this support if they want to use event management driver.
Thanks,
Jay
>
> thanks,
>
> greg k-h
On Wed, Oct 04, 2023 at 12:48:25PM +0000, Buddhabhatti, Jay wrote:
> Hi Greg,
>
> > -----Original Message-----
> > From: Greg KH <gregkh@linuxfoundation.org>
> > Sent: Wednesday, October 4, 2023 5:45 PM
> > To: Buddhabhatti, Jay <jay.buddhabhatti@amd.com>
> > Cc: Simek, Michal <michal.simek@amd.com>; Shah, Tanmay
> > <tanmay.shah@amd.com>; Manne, Nava kishore
> > <nava.kishore.manne@amd.com>; Levinsky, Ben <ben.levinsky@amd.com>;
> > robh@kernel.org; dhaval.r.shah@amd.com; marex@denx.de; Shaikh, Izhar
> > Ameer <izhar.ameer.shaikh@amd.com>; arnd@arndb.de;
> > ruanjinjie@huawei.com; mathieu.poirier@linaro.org; linux-
> > kernel@vger.kernel.org; linux-arm-kernel@lists.infradead.org
> > Subject: Re: [PATCH v2 4/6] drivers: soc: xilinx: Fix error message on SGI
> > registration failure
> >
> > On Wed, Oct 04, 2023 at 02:41:14AM -0700, Jay Buddhabhatti wrote:
> > > Failure to register SGI for firmware event notification is non-fatal
> > > error when feature is not supported by other modules such as Xen and
> > > TF-A. Add _info level log message for such special case.
> > >
> > > Also add XST_PM_INVALID_VERSION error code and map it to -EOPNOSUPP
> > > Linux kernel error code. If feature is not supported or EEMI API
> > > version is mismatch, firmware can return XST_PM_INVALID_VERSION = 4 or
> > > XST_PM_NO_FEATURE = 19 error code.
> > >
> > > Co-developed-by: Tanmay Shah <tanmay.shah@amd.com>
> > > Signed-off-by: Tanmay Shah <tanmay.shah@amd.com>
> > > Signed-off-by: Jay Buddhabhatti <jay.buddhabhatti@amd.com>
> > > ---
> > > drivers/firmware/xilinx/zynqmp.c | 4 +++-
> > > drivers/soc/xilinx/xlnx_event_manager.c | 7 ++++++-
> > > include/linux/firmware/xlnx-zynqmp.h | 1 +
> > > 3 files changed, 10 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/drivers/firmware/xilinx/zynqmp.c
> > > b/drivers/firmware/xilinx/zynqmp.c
> > > index 00ea2508f692..de09d77cf1e4 100644
> > > --- a/drivers/firmware/xilinx/zynqmp.c
> > > +++ b/drivers/firmware/xilinx/zynqmp.c
> > > @@ -92,6 +92,8 @@ static int zynqmp_pm_ret_code(u32 ret_status)
> > > return 0;
> > > case XST_PM_NO_FEATURE:
> > > return -ENOTSUPP;
> > > + case XST_PM_INVALID_VERSION:
> > > + return -EOPNOTSUPP;
> > > case XST_PM_NO_ACCESS:
> > > return -EACCES;
> > > case XST_PM_ABORT_SUSPEND:
> > > @@ -379,7 +381,7 @@ int zynqmp_pm_register_sgi(u32 sgi_num, u32 reset)
> > >
> > > ret = zynqmp_pm_invoke_fn(TF_A_PM_REGISTER_SGI, sgi_num, reset,
> > 0, 0,
> > > 0, NULL);
> > > - if (!ret)
> > > + if (ret != -EOPNOTSUPP && !ret)
> > > return ret;
> > >
> > > /* try old implementation as fallback strategy if above fails */
> > > diff --git a/drivers/soc/xilinx/xlnx_event_manager.c
> > > b/drivers/soc/xilinx/xlnx_event_manager.c
> > > index 38cfc161a713..bfa5cf4a0cab 100644
> > > --- a/drivers/soc/xilinx/xlnx_event_manager.c
> > > +++ b/drivers/soc/xilinx/xlnx_event_manager.c
> > > @@ -653,7 +653,12 @@ static int xlnx_event_manager_probe(struct
> > > platform_device *pdev)
> > >
> > > ret = zynqmp_pm_register_sgi(sgi_num, 0);
> > > if (ret) {
> > > - dev_err(&pdev->dev, "SGI %d Registration over TF-A failed with
> > %d\n", sgi_num, ret);
> > > + if (ret == -EOPNOTSUPP)
> > > + dev_info(&pdev->dev, "PM firmware event notification
> > not
> > > +supported\n");
> >
> > Why is this "info"? Why not error?
> [Jay] Because it is non-fatal error which will not cause kernel crash.
Neither do dev_err() messages :)
And something failed, this isn't just "here's some nice information", it
is "something went wrong" which should be a warning or error in my
opinion.
> > And what is userspace supposed to do with this information if it sees it?
> [Jay] This message indicates that SGI registration is not supported by other modules like TF-A or Xen. So user need to use TF-A with this support if they want to use event management driver.
So perhaps say that here too? Otherwise how are they supposed to know
that?
thanks,
greg k-h
© 2016 - 2025 Red Hat, Inc.