linux-kernel - Re: [PATCH v4 5/5] mfd: tqmx86: add I2C IRQ support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20241016130715.GA1152434@google.com>
Date: Wed, 16 Oct 2024 14:07:15 +0100
From: Lee Jones <lee@...nel.org>
To: Matthias Schiffer <matthias.schiffer@...tq-group.com>
Cc: linux@...tq-group.com, linux-kernel@...r.kernel.org,
	Gregor Herburger <gregor.herburger@...group.com>
Subject: Re: [PATCH v4 5/5] mfd: tqmx86: add I2C IRQ support

On Wed, 16 Oct 2024, Matthias Schiffer wrote:

> On Tue, 2024-10-15 at 12:38 +0100, Lee Jones wrote:
> > 
> > On Mon, 07 Oct 2024, Matthias Schiffer wrote:
> > 
> > > From: Gregor Herburger <gregor.herburger@...group.com>
> > > 
> > > The i2c-ocores controller can run in interrupt mode on tqmx86 modules.
> > > Add module parameter to allow configuring the IRQ number, similar to the
> > > handling of the GPIO IRQ.
> > > 
> > > Signed-off-by: Gregor Herburger <gregor.herburger@...group.com>
> > > Signed-off-by: Matthias Schiffer <matthias.schiffer@...tq-group.com>
> > > ---
> > > 
> > > v2: improve module parameter description (was patch 4/4)
> > > v3: replace IRQ 0 resource with an empty placeholder to simplify error handling
> > > v4: no changes
> > > 
> > >  drivers/mfd/tqmx86.c | 20 +++++++++++++++++++-
> > >  1 file changed, 19 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/mfd/tqmx86.c b/drivers/mfd/tqmx86.c
> > > index e444fcd2a3e9d..057d035b71d33 100644
> > > --- a/drivers/mfd/tqmx86.c
> > > +++ b/drivers/mfd/tqmx86.c
> > > @@ -50,6 +50,7 @@
> > >  #define TQMX86_REG_IO_EXT_INT_9			2
> > >  #define TQMX86_REG_IO_EXT_INT_12		3
> > >  #define TQMX86_REG_IO_EXT_INT_MASK		0x3
> > > +#define TQMX86_REG_IO_EXT_INT_I2C_SHIFT		0
> > >  #define TQMX86_REG_IO_EXT_INT_GPIO_SHIFT	4
> > >  #define TQMX86_REG_SAUC		0x17
> > >  
> > > @@ -60,7 +61,16 @@ static uint gpio_irq;
> > >  module_param(gpio_irq, uint, 0);
> > >  MODULE_PARM_DESC(gpio_irq, "GPIO IRQ number (valid parameters: 7, 9, 12)");
> > >  
> > > -static const struct resource tqmx_i2c_soft_resources[] = {
> > > +static uint i2c_irq;
> > > +module_param(i2c_irq, uint, 0);
> > > +MODULE_PARM_DESC(i2c_irq, "I2C IRQ number (valid parameters: 7, 9, 12)");
> > > +
> > > +static struct resource tqmx_i2c_soft_resources[] = {
> > > +	/*
> > > +	 * Placeholder for IRQ resource - must come first to be filled in by the
> > > +	 * probe function.
> > > +	 */
> > > +	{},
> > 
> > Having a NULLed entry in the first slot doesn't sit well with me at all.
> > 
> > In order for us to avoid wasting memory, it would be better to place the
> > entry at the end of the array with a blank entry:
> > 
> >   DEFINE_RES_IRQ(0),
> > 
> > Later comes the matching code which updates the 0 value to something sane.
> > 
> > Before you call to the add the devices, check to see if the value has
> > changed.  If it hasn't, deprecate num_resources, effectively masking the
> > last entry in the array.  Then when platform_device_add_resources()
> > comes to duplicate the array, it will only copy the relevant entries.
> 
> I chose my current solution because it resulted in more simple and maintainable code:
> 
> - No dynamic array access, the IRQ resource is always written to index 0

Which is fragile (even with the comment).  If you're going to directly
index array elements, please do so with a unique identifier rather than
relying on it happening to reside in the first.

> - No surprises regarding num_resources, it is always equal to ARRAY_SIZE(resources)

No surprises, but sometimes it's incorrect.

> It also allowed to make all mfd_cell const; to make num_resources modifyable, either the const would
> need to be removed, or each mfd_cell would need to be copied to a stack variable in the probe
> function.
> 
> In my opinion, these benefits outweigh the 2*64 bytes saved for the additional resources allocated
> by platform_device_add_resources() - and 128 bytes doesn't seem significant at all, in particular
> considering that this driver is used on x86_64 only.

But doesn't outweigh my disliking for placing a NULL element at the
start of the array.  If you _must_ do something like this, please place
it at the end of the array.

> > 
> > >  	DEFINE_RES_IO(TQMX86_IOBASE_I2C, TQMX86_IOSIZE_I2C),
> > >  };
> > >  
> > > @@ -263,6 +273,14 @@ static int tqmx86_probe(struct platform_device *pdev)
> > >  	ocores_platform_data.clock_khz = tqmx86_board_id_to_clk_rate(dev, board_id);
> > >  
> > >  	if (i2c_det == TQMX86_REG_I2C_DETECT_SOFT) {
> > > +		if (i2c_irq) {
> > > +			err = tqmx86_setup_irq(dev, "I2C", i2c_irq, io_base,
> > > +					       TQMX86_REG_IO_EXT_INT_I2C_SHIFT);
> > > +			if (!err)
> > > +				/* Assumes the IRQ resource placeholder is first */
> > > +				tqmx_i2c_soft_resources[0] = DEFINE_RES_IRQ(i2c_irq);
> > > +		}
> > > +
> > >  		err = devm_mfd_add_devices(dev, PLATFORM_DEVID_NONE,
> > >  					   tqmx86_i2c_soft_dev,
> > >  					   ARRAY_SIZE(tqmx86_i2c_soft_dev),
> > > -- 
> > > TQ-Systems GmbH | Mühlstraße 2, Gut Delling | 82229 Seefeld, Germany
> > > Amtsgericht München, HRB 105018
> > > Geschäftsführer: Detlef Schneider, Rüdiger Stahl, Stefan Schneider
> > > https://www.tq-group.com/
> > > 
> > 
> 
> -- 
> TQ-Systems GmbH | Mühlstraße 2, Gut Delling | 82229 Seefeld, Germany
> Amtsgericht München, HRB 105018
> Geschäftsführer: Detlef Schneider, Rüdiger Stahl, Stefan Schneider
> https://www.tq-group.com/

-- 
Lee Jones [李琼斯]