linux-kernel - Re: [PATCH] sd: make ->no_write_same independent of reported ->max_ws

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: <87lgvteaz0.fsf@gmail.com>
Date:   Tue, 06 Dec 2016 09:08:19 +0100
From:   Nicolai Stange <nicstange@...il.com>
To:     "Martin K. Petersen" <martin.petersen@...cle.com>
Cc:     Nicolai Stange <nicstange@...il.com>, Jens Axboe <axboe@...nel.dk>,
        "James E.J. Bottomley" <jejb@...ux.vnet.ibm.com>,
        Shaun Tancheff <shaun.tancheff@...gate.com>,
        Chaitanya Kulkarni <chaitanya.kulkarni@...t.com>,
        Christoph Hellwig <hch@...radead.org>,
        linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] sd: make ->no_write_same independent of reported ->max_ws_blocks

Hello Martin,

"Martin K. Petersen" <martin.petersen@...cle.com> writes:

>>>>>> "Nicolai" == Nicolai Stange <nicstange@...il.com> writes:
> Nicolai> Due to reported problems with Write Same on ATA devices, commit
> Nicolai> 0ce1b18c42a5 ("libata: Some drives failing on SCT Write Same")
> Nicolai> strived to report non-support for Write Same on non-zoned ATA
> Nicolai> devices.
>
> Nicolai> However, due to the following control flow in
> Nicolai> sd_config_write_same() this doesn't always take effect, namely
> Nicolai> if the ->max_ws_blocks as set in the by the ATA Identify Device
> Nicolai> exceeds SD_WS10_BLOCKS:
>
> I'd much prefer for libata to set no_write_same = 1 for non-ZAC devices.

Or just try it once and let the sd layer, i.e. sd_done(), disable it
once a ILLEGAL COMMAND OPCODE is reported. This works right now and as
you said below, calling code must cope gracefully with a failing Write
Same anyway (which doesn't work right now).

>
> Older SCSI devices have no way to explicitly report that WRITE SAME is
> supported. So the heuristic is the way it is to permit trying WRITE SAME
> unless no_write_same has been set by the device driver.

Ok, I didn't see that there might be a heuristic going on.

I've got a couple of questions about this, but they're mainly out of
curiosity. So feel free to ignore them.

1.) Do these older SCSI devices have a way to report ->max_ws_blocks?
    Because otherwise the heuristic would not work?
    Or is it set speculatively somewhere?

2.) If so, what about such older devices having
    0 < ->max_ws_blocks < SD_MAX_WS10_BLOCKS ?
    Wouldn't these also be suitable candidates for trying that
    heuristic on?

3.) Those older devices that have ->max_ws_blocks > SD_MAX_WS10_BLOCKS
    but ->ws16 == ->ws10 == 0, i.e. the heuristicated ones would
    always be given WRITE_SAME, not WRITE_SAME_16 commands?
    C.f. sd_setup_write_same_cmnd(): if ->ws16 is not set, do
    WRITE_SAME. Isn't this a little bit odd given that the reported 
    ->max_ws_blocks would be greater than SD_MAX_WS10_BLOCKS?
    Ok, given that these devices are older anyway, WRITE_SAME seems
    like the obvious choice to be made over WRITE_SAME_16. Which
    brings me back to question 2.).

    The answer to this question would possibly affect ATA devices with
    this heuristic going on as well: according to ata_scsiop_maint_in(),
    they would only support WRITE_SAME_16, but not WRITE_SAME.

    Heck, this is perhaps the reason why I'm seeing those errors this
    commit 0ce1b18c42a5 ("libata: Some drives failing on SCT Write
    Same") effectively turns the heuristics for my ATA device on,
    i.e. unsets ->ws16, resulting in WRITE_SAME's which are unsupported
    by libata-scsi, c.f. ata_get_xlat_func()...

>
> Nicolai> Since commit e73c23ff736e ("block: add async variant of
> Nicolai> blkdev_issue_zeroout"), blkdev_issue_zeroout() got a little bit
> Nicolai> more sensitive towards failing Write Sames on devices that
> Nicolai> claim to support them and this results in messages like
>
> That's something that needs to be addressed. blkdev_issue_zeroout() must
> cope with WRITE SAME failing and fall back to a manual zeroout.

That's very useful information! So this commit really needs a fixup in
either way.


Thank you!

Nicolai