Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
IronPort-SDR: JEo/yV0WM4soNmUhef3oyE5YY0NLTPN8yHv08O71Vg+Id23CwcjWCPXD8dtJ4OPK7nvMG1xbRa
 RKJZZiovLP+ahmgXocDNtnciKZFBDbi10ZT2k88PFTDwfoDs1gKtMgPuW7IbLsJOHIZoKFtSdK
 FeZfPw8uM9V02FsrPyUW0GED9DVwAEuSx9ogtK07VJXJnHixu81UlIsI4oLUTp2+UOfMN68C0m
 scA8ehpz4qkUViE/hIOZNyFQlJ59qJ5N9pdQ1DJ5iaRY+ELr8aAHNvqOL1T81QovkADO/Ix74U
 I7s=
IronPort-SDR: Lzwx2wwI0kpb+QCIZwVAvoKhpA/GSLo3zlDG5dLyZT+5nLHnB08dxuVBLMiH+p0BJQIYRZxW1S
 1YHET0k+x1DdDszaZ8kVboIW+38oJDAuBGqLeK7bAHrAb9yQofFgTD0ZeHYsE08jibAsT/9onU
 K+89abLpRhV42rrk5hlTX+9ouKAJnnrVjV7a0JGZNJKd2B309zMqNvuBQlwXElmew03fWimTYF
 FmWOCIC3EZrRgI01MuLUKclniFUzFfJvZ1ahHx9z+9r7S5X1p/klLE2C1NNKLWVUmnjeaj4Ntw
 sSE=
WDCIronportException: Internal
Message-ID: <f77c7872-3711-2216-c17c-e23591f745c7@opensource.wdc.com>
Date:   Wed, 21 Dec 2022 17:31:59 +0900
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.6.0
Subject: Re: [PATCH] scsi: libsas: Grab the host lock in
 sas_ata_device_link_abort()
To:     Jason Yan <yanaijie@huawei.com>,
        John Garry <john.g.garry@oracle.com>,
        Xingui Yang <yangxingui@huawei.com>, jejb@linux.ibm.com,
        martin.petersen@oracle.com, niklas.cassel@wdc.com
Cc:     linux-scsi@vger.kernel.org, linux-kernel@vger.kernel.org,
        linuxarm@huawei.com, prime.zeng@hisilicon.com,
        kangfenglong@huawei.com
References: <20221220125349.45091-1-yangxingui@huawei.com>
 <4ec9dbed-1758-d6b4-dc1d-ac42e8c22731@oracle.com>
 <c8387766-2ca0-51f3-e332-71492b13e5c1@opensource.wdc.com>
 <7347d117-6e0b-dd18-90a8-25685f757689@huawei.com>
 <4ff0ca00-31f5-2867-ff59-cecb5d6d1048@opensource.wdc.com>
 <755d7a9c-427e-024a-8509-449ebc5a00e6@huawei.com>
Content-Language: en-US
From:   Damien Le Moal <damien.lemoal@opensource.wdc.com>
Organization: Western Digital Research
In-Reply-To: <755d7a9c-427e-024a-8509-449ebc5a00e6@huawei.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On 12/21/22 15:35, Jason Yan wrote:
> On 2022/12/21 11:59, Damien Le Moal wrote:
>> On 2022/12/21 11:42, Jason Yan wrote:
>>> On 2022/12/21 8:36, Damien Le Moal wrote:
>>>> On 2022/12/20 23:59, John Garry wrote:
>>>>> On 20/12/2022 12:53, Xingui Yang wrote:
>>>>>> Grab the host lock in sas_ata_device_link_abort() before calling
>>>>>
>>>>> This is should be the ata port lock, right? I know that the ata
>>>>> comments
>>>>> say differently.
>>>>>
>>>>>> ata_link_abort(), as the comment in ata_link_abort() mentions.
>>>>>>
>>>>>
>>>>> Can you please add a fixes tag?
>>>>>
>>>>>> Signed-off-by: Xingui Yang <yangxingui@huawei.com>
>>>>>
>>>>> Reviewed-by: John Garry <john.g.garry@oracle.com>
>>>>>
>>>>>> ---
>>>>>> =C2=A0=C2=A0=C2=A0 drivers/scsi/libsas/sas_ata.c | 3 +++
>>>>>> =C2=A0=C2=A0=C2=A0 1 file changed, 3 insertions(+)
>>>>>>
>>>>>> diff --git a/drivers/scsi/libsas/sas_ata.c
>>>>>> b/drivers/scsi/libsas/sas_ata.c
>>>>>> index f7439bf9cdc6..4f2017b21e6d 100644
>>>>>> --- a/drivers/scsi/libsas/sas_ata.c
>>>>>> +++ b/drivers/scsi/libsas/sas_ata.c
>>>>>> @@ -889,7 +889,9 @@ void sas_ata_device_link_abort(struct
>>>>>> domain_device *device, bool force_reset)
>>>>>> =C2=A0=C2=A0=C2=A0 {
>>>>>> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 struct ata_port *ap =3D=
 device->sata_dev.ap;
>>>>>> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 struct ata_link *link =3D=
 &ap->link;
>>>>>> +=C2=A0=C2=A0=C2=A0 unsigned long flags;
>>>>>> =C2=A0=C2=A0=C2=A0 +=C2=A0=C2=A0=C2=A0 spin_lock_irqsave(ap->lock,=
 flags);
>>>>>> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 device->sata_dev.fis[2]=
 =3D ATA_ERR | ATA_DRDY; /* tf status */
>>>>>> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 device->sata_dev.fis[3]=
 =3D ATA_ABORTED; /* tf error */
>>>>>> =C2=A0=C2=A0=C2=A0 @@ -897,6 +899,7 @@ void sas_ata_device_link_ab=
ort(struct
>>>>>> domain_device *device, bool force_reset)
>>>>>> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 if (force_reset)
>>>>>> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
 link->eh_info.action |=3D ATA_EH_RESET;
>>>>>> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 ata_link_abort(link);
>>>>
>>>> Really need to add lockdep annotations in libata to avoid/catch such
>>>> bugs...
>>>> Will work on that.
>>>
>>> Actually in libata there are many places calling ata_link_abort() not
>>> holding port lock. And some places are holding the real host
>>> lock(ata_host->lock) while calling ata_link_abort(). So if you add th=
e
>>> lockdep annotations, there may be too many warnings. If these are rea=
l
>>> issues, we should fix them first.
>>
>> libata-EH does most of its work without the port lock held because by
>> the time
>> we get EH started, we are guaranteed to be idle with no commands in
>> flight. That
>> is why the calls you mention look like "bugs" but are not.
>=20
> What about the interrupt handler such as ahci_error_intr()? I didn't se=
e
> the callers hold the port lock too. Do they need the port lock?

It looks like it is missing for ahci_thunderx_irq_handler() but that one
takes the host lock. Same for xgene_ahci_irq_intr(), again no port lock
but host lock taken. And again for ahci_single_level_irq_intr() for the
non MSI case. For modern MSI adapters, the port lock is taken in

For other cases, ahci_multi_irqs_intr_hard) takes the port lock.

So it looks like ahci_port_intr() needs to take the lock and some
cleanups overall (the host lock should not be necessary in the command
path. But nobody seems to have issues with the "bad" cases... Probably
because they are not mainstream adapters.

Definitely some work needed here.

--=20
Damien Le Moal
Western Digital Research