[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <871q5b3wgr.ffs@tglx>
Date: Wed, 05 Jun 2024 23:34:28 +0200
From: Thomas Gleixner <tglx@...utronix.de>
To: lakshmi.sowjanya.d@...el.com, giometti@...eenne.com, corbet@....net,
linux-kernel@...r.kernel.org, linux-doc@...r.kernel.org
Cc: gregkh@...uxfoundation.org, andriy.shevchenko@...ux.intel.com,
eddie.dong@...el.com, christopher.s.hall@...el.com, pandith.n@...el.com,
subramanian.mohan@...el.com, thejesh.reddy.t.r@...el.com,
lakshmi.sowjanya.d@...el.com
Subject: Re: [PATCH v9 1/3] pps: generators: Add PPS Generator TIO Driver
On Wed, Jun 05 2024 at 21:05, lakshmi.sowjanya.d@...el.com wrote:
> +static ssize_t enable_store(struct device *dev, struct device_attribute *attr, const char *buf,
> + size_t count)
> +{
> + struct pps_tio *tio = dev_get_drvdata(dev);
> + bool enable;
> + int err;
> +
> + if (!timekeeping_clocksource_has_base(CSID_X86_ART)) {
> + dev_err(dev, "PPS cannot be used as clock is not related to ART");
dev_err_once() if at all
> + return -EPERM;
Why -EPERM? This has nothing to do with permissions.
ENODEV or ENOTSUPPORTED perhaps.
> +static ssize_t enable_show(struct device *dev, struct device_attribute *devattr, char *buf)
> +{
> + struct pps_tio *tio = dev_get_drvdata(dev);
> + u32 ctrl;
> +
> + ctrl = pps_tio_read(tio, TIOCTL);
> + ctrl &= TIOCTL_EN;
Why reading the hardware instead of simply using tio->enabled?
> +
> + return sysfs_emit(buf, "%u\n", ctrl);
> +}
Thanks,
tglx
Powered by blists - more mailing lists