Re: [PATCH v2 2/3] fpga manager: framework core

[atull <atull@opensource.altera.com>]

On Sat, 6 Dec 2014, Grant Likely wrote:

> >> +int fpga_mgr_write(struct fpga_manager *mgr, const char *buf, size_t count)
> >> +{
> >> +     int ret;
> >> +
> >> +     if (test_and_set_bit_lock(FPGA_MGR_BUSY, &mgr->flags))
> >> +             return -EBUSY;
> >> +
> >> +     dev_info(mgr->dev, "writing buffer to %s\n", mgr->name);
> >> +
> >> +     ret = __fpga_mgr_write(mgr, buf, count);
> >> +     clear_bit_unlock(FPGA_MGR_BUSY, &mgr->flags);
> >> +
> >> +     return ret;
> >> +}
> >> +EXPORT_SYMBOL_GPL(fpga_mgr_write);
> >
> > Is the EBUSY -- userspace please try again, but you don't know when to
> > try again -- right interface? I mean, normally kernel would wait, so
> > that userland does not have to poll?
> 
> Custom locking schemes are just wrong. A mutex is the right thing to
> do here and then an -EBUSY isn't required.
> 

I've changed it to a mutex in the next version.

> >
> >> +static ssize_t fpga_mgr_firmware_store(struct device *dev,
> >> +                                    struct device_attribute *attr,
> >> +                                    const char *buf, size_t count)
> >> +{
> >> +     struct fpga_manager *mgr = dev_get_drvdata(dev);
> >> +     unsigned int len;
> >> +     char image_name[NAME_MAX];
> 
> Hard coding a string length is a warning sign. That is the sort of
> thing that can memdup() or strdup() can handle.
> 

OK, I'll fix it using kstrdup() in v6.

Alan