Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
MIME-Version: 1.0
References: <CAMhUBjkVME8D5KsHvT=uddBsW_Bh6wr7qeXS=UpQD4LgPmHffQ@mail.gmail.com>
 <05433153-0424-ab66-1573-993d0490c5bc@opensource.wdc.com> <CAMhUBj=JsiHnnQzrqPKzA=Z2+589Ju_HE0cFKyon58Fk0waeAQ@mail.gmail.com>
 <3045fde1-2d5b-e731-ef18-6cf3f0987259@opensource.wdc.com>
In-Reply-To: <3045fde1-2d5b-e731-ef18-6cf3f0987259@opensource.wdc.com>
From:   Zheyu Ma <zheyuma97@gmail.com>
Date:   Wed, 20 Apr 2022 10:21:14 +0800
Message-ID: <CAMhUBjnG75XANF1E5KUyigKSmhxzvJytfapieC6_jpKdiFwzSg@mail.gmail.com>
Subject: Re: [BUG] ata: pata_marvell: Warning when probing the module
To:     Damien Le Moal <damien.lemoal@opensource.wdc.com>
Cc:     Sergey Shtylyov <s.shtylyov@omp.ru>, linux-ide@vger.kernel.org,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Apr 13, 2022 at 11:42 AM Damien Le Moal
<damien.lemoal@opensource.wdc.com> wrote:
>
> On 4/12/22 15:34, Zheyu Ma wrote:
> > On Mon, Apr 11, 2022 at 7:53 AM Damien Le Moal
> > <damien.lemoal@opensource.wdc.com> wrote:
> >>
> >> On 4/10/22 15:30, Zheyu Ma wrote:
> >>> Hello,
> >>>
> >>> I found a bug in the pata_marvell module.
> >>> When probing the driver, it seems to trigger the error path and
> >>> executes the function marvell_cable_detect(), but the
> >>> 'ap->ioaddr.bmdma_addr' is not initialized, which causes a warning.
> >>
> >> I do not have this hardware so I cannot debug this. Please debug it and
> >> send a patch. bmdma_addr is normally set in ata_pci_bmdma_init(), but some
> >> drivers set it manually in their probe functions. No idea about the
> >> marvell driver, I have not checked it.
> >
> > To be honest I don't have a good solution to this problem, because
> > other drivers don't have similar behavior. The marvell driver doesn't
> > even initialize 'bmdma_addr' before calling 'cable_detect'.
>
> Then this is the bug that needs to be fixed, no ?
>
> > So a simple idea I have is to check if 'bmdma_addr' is 0 before
> > reading it and if so return the error code ATA_CBL_NONE.
>
> And if indeed, even after it is initialized it is still 0, then yes, this
> change seems sensible.

Sorry for the late reply, I found the root cause of this issue.
The marvell driver execute the ata_pci_bmdma_init() function, but the
driver just returned at the following code snippet.

if (pci_resource_start(pdev, 4) == 0) {
      ata_bmdma_nodma(host, "BAR4 is zero");
      return;
}

So the driver didn't initialize the 'bmdma_addr' but used it in the
cable_detect() function.
It seems that the problem is caused by the hardware, is this a bug
that we should fix?

Zheyu Ma


Zheyu Ma