Re: [PATCH V5 02/11] misc: xilinx-sdfec: add core driver

[Greg KH]

On Sun, Jun 09, 2019 at 06:48:31PM +0000, Dragan Cvetic wrote:
> 
> 
> > -----Original Message-----
> > From: Greg KH [mailto:gregkh@xxxxxxxxxxxxxxxxxxx]
> > Sent: Sunday 9 June 2019 12:23
> > To: Dragan Cvetic <draganc@xxxxxxxxxx>
> > Cc: arnd@xxxxxxxx; Michal Simek <michals@xxxxxxxxxx>; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; robh+dt@xxxxxxxxxx;
> > mark.rutland@xxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Derek Kiernan <dkiernan@xxxxxxxxxx>
> > Subject: Re: [PATCH V5 02/11] misc: xilinx-sdfec: add core driver
> > 
> > On Sun, Jun 09, 2019 at 01:04:07AM +0100, Dragan Cvetic wrote:
> > > Implement a platform driver that matches with xlnx,
> > > sd-fec-1.1 device tree node and registers as a character
> > > device, including:
> > > - SD-FEC driver binds to sdfec DT node.
> > > - creates and initialise an initial driver dev structure.
> > > - add the driver in Linux build and Kconfig.
> > >
> > > Tested-by: Dragan Cvetic <dragan.cvetic@xxxxxxxxxx>
> > > Signed-off-by: Derek Kiernan <derek.kiernan@xxxxxxxxxx>
> > > Signed-off-by: Dragan Cvetic <dragan.cvetic@xxxxxxxxxx>
> > > ---
> > >  drivers/misc/Kconfig        |  12 +++++
> > >  drivers/misc/Makefile       |   1 +
> > >  drivers/misc/xilinx_sdfec.c | 118 ++++++++++++++++++++++++++++++++++++++++++++
> > >  3 files changed, 131 insertions(+)
> > >  create mode 100644 drivers/misc/xilinx_sdfec.c
> > >
> > > diff --git a/drivers/misc/Kconfig b/drivers/misc/Kconfig
> > > index 6b0417b..319a6bf 100644
> > > --- a/drivers/misc/Kconfig
> > > +++ b/drivers/misc/Kconfig
> > > @@ -471,6 +471,18 @@ config PCI_ENDPOINT_TEST
> > >             Enable this configuration option to enable the host side test driver
> > >             for PCI Endpoint.
> > >
> > > +config XILINX_SDFEC
> > > +	tristate "Xilinx SDFEC 16"
> > > +	help
> > > +	  This option enables support for the Xilinx SDFEC (Soft Decision
> > > +	  Forward Error Correction) driver. This enables a char driver
> > > +	  for the SDFEC.
> > > +
> > > +	  You may select this driver if your design instantiates the
> > > +	  SDFEC(16nm) hardened block. To compile this as a module choose M.
> > > +
> > > +	  If unsure, say N.
> > > +
> > >  config MISC_RTSX
> > >  	tristate
> > >  	default MISC_RTSX_PCI || MISC_RTSX_USB
> > > diff --git a/drivers/misc/Makefile b/drivers/misc/Makefile
> > > index b9affcd..0cb3546 100644
> > > --- a/drivers/misc/Makefile
> > > +++ b/drivers/misc/Makefile
> > > @@ -59,3 +59,4 @@ obj-$(CONFIG_OCXL)		+= ocxl/
> > >  obj-y				+= cardreader/
> > >  obj-$(CONFIG_PVPANIC)   	+= pvpanic.o
> > >  obj-$(CONFIG_HABANA_AI)		+= habanalabs/
> > > +obj-$(CONFIG_XILINX_SDFEC)	+= xilinx_sdfec.o
> > > diff --git a/drivers/misc/xilinx_sdfec.c b/drivers/misc/xilinx_sdfec.c
> > > new file mode 100644
> > > index 0000000..75cc980
> > > --- /dev/null
> > > +++ b/drivers/misc/xilinx_sdfec.c
> > > @@ -0,0 +1,118 @@
> > > +// SPDX-License-Identifier: GPL-2.0
> > > +/*
> > > + * Xilinx SDFEC
> > > + *
> > > + * Copyright (C) 2019 Xilinx, Inc.
> > > + *
> > > + * Description:
> > > + * This driver is developed for SDFEC16 (Soft Decision FEC 16nm)
> > > + * IP. It exposes a char device which supports file operations
> > > + * like  open(), close() and ioctl().
> > > + */
> > > +
> > > +#include <linux/miscdevice.h>
> > > +#include <linux/io.h>
> > > +#include <linux/interrupt.h>
> > > +#include <linux/kernel.h>
> > > +#include <linux/module.h>
> > > +#include <linux/of_platform.h>
> > > +#include <linux/poll.h>
> > > +#include <linux/slab.h>
> > > +#include <linux/clk.h>
> > > +
> > > +static int xsdfec_ndevs;
> > 
> > You should use an idr for this, not just a number you bump up and down.
> > This will not work properly at all.
> > 
> > Think about this situation:
> > 	probe device 0
> > 	xsdfec_ndevs = 1
> > 	probe device 1
> > 	xsdfec_ndevs = 2
> > 	remove device 0
> > 	xsdfec_ndevs = 0
> > 	probe another device
> > 	misc device fails due to duplicate name.
> >
> 
> My bad.
> I can use idr, but couldn't be better optimized code if use simple mutex to protect the variable.

mutex does not protect from this at all, it's a logic bug.  Think about
adding 5 devices and then removing the 2nd one.  What is the number
assigned to the new device that is added afterward?

And you need a mutex for the idr anyway, if you are touching it in a non
probe/release callback way (those are already serialized by the bus
lock).

thanks,

greg k-h