linux-kernel - Re: [char-misc-next] mei: simplify error handling via devres function.

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]

Date:   Mon, 23 Jan 2017 16:48:53 +0200
From:   Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
To:     "Winkler, Tomas" <tomas.winkler@...el.com>,
        Andy Shevchenko <andy.shevchenko@...il.com>
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "Usyskin, Alexander" <alexander.usyskin@...el.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [char-misc-next] mei: simplify error handling via devres
 function.

On Sat, 2017-01-21 at 10:12 +0000, Winkler, Tomas wrote:

> > 
> > > -struct mei_device *mei_txe_dev_init(struct pci_dev *pdev)
> > > +struct mei_device *devm_mei_txe_init(struct pci_dev *pdev)
> > 
> > Ditto.
> > 
> > >  end:
> > > +       pci_set_drvdata(pdev, NULL);
> > 
> > Not needed.
> 
> Please explain, we rely on pci_get_drvdata() returning NULL when
> unregistered. 

PCI core will take care about this one. Actually device core for any of
user of struct device.
See __device_release_driver() for the details.

> > 
> > > -       free_irq(pdev->irq, dev);
> > > +       devm_free_irq(&pdev->dev, pdev->irq, dev);
> > >         pci_disable_msi(pdev);
> > 
> > All three not needed
> 
> I believe we need it on suspend as we are going over  irq request
> again in resume.  Please provide more info you if you still insist. 

Ah, sorry, I missed that these are suspend/resume hooks.

So, Can you elaborate a bit why you need to disable interrupts during
system suspend?

(Basically in this case better not to use devm_request_*irq() at all)

> > 
> > >         return 0;
> > > @@ -75,22 +64,22 @@ static int mei_txe_probe(struct pci_dev *pdev,
> > > const struct pci_device_id *ent)  {
> > >         struct mei_device *dev;
> > >         struct mei_txe_hw *hw;
> > > +       const int mask = BIT(SEC_BAR) | BIT(BRIDGE_BAR);
> > 
> > First line?
> 
> Please be more verbose.

Use reversed tree for definition block.

The longest lines with the assignment = first;
Then lines without assignment;
Then return code variable;

Flags for spin_lock -- depends.

> > 
> > > +       memcpy(hw->mem_addr, pcim_iomap_table(pdev),
> > > + sizeof(hw->mem_addr));
> > 
> > Why?
> > It is kept by PCI core, you don't need a copy.
> 
> There is no simple accessor for that, it's easier to copy the two
> dwords then going over the function calls. 

I'm not sure you need a copy. That function call just return the pointer
to the table.

I remember 8250_pci used to have similar approach, now it's using
whatever is kept by PCI core.

It's less error prone.


> > > @@ -256,7 +210,7 @@ static int mei_txe_pci_suspend(struct device
> > > *device)
> > > -       free_irq(pdev->irq, dev);
> > > +       devm_free_irq(&pdev->dev, pdev->irq, dev);
> > >         pci_disable_msi(pdev);
> > 
> > All are redundant.

Yeah, same clarification as for above case with system sleep.

-- 
Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
Intel Finland Oy