linux-kernel - simplify gendisk lookup and remove struct block

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 29 Oct 2020 15:58:23 +0100
From:   Christoph Hellwig <hch@....de>
To:     Jens Axboe <axboe@...nel.dk>
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "Rafael J. Wysocki" <rafael@...nel.org>,
        Denis Efremov <efremov@...ux.com>,
        "David S. Miller" <davem@...emloft.net>,
        Song Liu <song@...nel.org>, Al Viro <viro@...iv.linux.org.uk>,
        Finn Thain <fthain@...egraphics.com.au>,
        Michael Schmitz <schmitzmic@...il.com>,
        linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-ide@...r.kernel.org, linux-raid@...r.kernel.org,
        linux-scsi@...r.kernel.org, linux-m68k@...ts.linux-m68k.org
Subject: simplify gendisk lookup and remove struct block_device aliases v4

Hi all,

this series removes the annoying struct block_device aliases, which can
happen for a bunch of old floppy drivers (and z2ram).  In that case
multiple struct block device instances for different dev_t's can point
to the same gendisk, without being partitions.  The cause for that
is the probe/get callback registered through blk_register_regions.

This series removes blk_register_region entirely, splitting it it into
a simple xarray lookup of registered gendisks, and a probe callback
stored in the major_names array that can be used for modprobe overrides
or creating devices on demands when no gendisk is found.  The old
remapping is gone entirely, and instead the 4 remaining drivers just
register a gendisk for each operating mode.  In case of the two drivers
that have lots of aliases that is done on-demand using the new probe
callback, while for the other two I simply register all at probe time
to keep things simple.

Note that the m68k drivers are compile tested only.

Changes since v3:
 - keep kobj_map for char dev lookup for now, as the testbot found
   some very strange and unexplained regressions, so I'll get back to
   this later separately
 - fix a commit message typo

Changes since v2:
 - fix a wrong variable passed to ERR_PTR in the floppy driver
 - slightly adjust the del_gendisk cleanups to prepare for the next
   series touching this area

Changes since v1:
 - add back a missing kobject_put in the cdev code
 - improve the xarray delete loops