linux-kernel - Re: [PATCH] libata: Use per port sync for detach

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, 18 May 2020 16:21:18 +0100
From:   John Garry <john.garry@...wei.com>
To:     Kai-Heng Feng <kai.heng.feng@...onical.com>
CC:     Jens Axboe <axboe@...nel.dk>,
        "open list:LIBATA SUBSYSTEM (Serial and Parallel ATA drivers)" 
        <linux-ide@...r.kernel.org>,
        open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] libata: Use per port sync for detach

On 18/05/2020 10:06, John Garry wrote:
> On 15/05/2020 18:48, Kai-Heng Feng wrote:
>>> 841]  ret_from_fork+0x10/0x1c
>>> [   28.393400] ---[ end trace 9972785c7052048f ]---
>>> [   28.435826] ahci 0000:b4:03.0: SSS flag set, parallel bus scan 
>>> disabled
>> Can you please test the following patch:
>>
>> diff --git a/drivers/ata/libata-eh.c b/drivers/ata/libata-eh.c
>> index 474c6c34fe02..51ee0cc4d414 100644
>> --- a/drivers/ata/libata-eh.c
>> +++ b/drivers/ata/libata-eh.c
>> @@ -3583,8 +3583,10 @@ int ata_eh_recover(struct ata_port *ap, 
>> ata_prereset_fn_t prereset,
>>          rc = 0;
>>          /* if UNLOADING, finish immediately */
>> -       if (ap->pflags & ATA_PFLAG_UNLOADING)
>> +       if (ap->pflags & ATA_PFLAG_UNLOADING) {
>> +               ap->pflags |= ATA_PFLAG_UNLOADED;
>>                  goto out;
>> +       }
>>
>> It's only compile-tested, many drivers panic with 
>> CONFIG_DEBUG_TEST_DRIVER_REMOVE enabled, so the system I have can't 
>> even boot properly:(
> 

According to the comment for async_synchronize_cookie(), we sync upto 
(but excluding) the cookie:

/**
  * async_synchronize_cookie - synchronize asynchronous function calls 
with cookie checkpointing
  *
  * This function waits until all asynchronous function calls prior to 
@cookie
  * have been done.

So maybe it should be:

+	for (i = 0; i < host->n_ports; i++)
+		async_synchronize_cookie(host->ports[i]->cookie + 1);

That is how other callsites use this API, and that change resolves the 
WARN for me.

Thanks,
John