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-next>] [day] [month] [year] [list]
Date:   Thu, 10 Dec 2020 09:18:32 -0800
From:   ira.weiny@...el.com
To:     Thomas Gleixner <tglx@...utronix.de>,
        Andrew Morton <akpm@...ux-foundation.org>
Cc:     Ira Weiny <ira.weiny@...el.com>,
        Joonas Lahtinen <joonas.lahtinen@...ux.intel.com>,
        Matthew Wilcox <willy@...radead.org>,
        Christoph Hellwig <hch@...radead.org>,
        Eric Biggers <ebiggers@...nel.org>,
        Dan Williams <dan.j.williams@...el.com>,
        Al Viro <viro@...iv.linux.org.uk>,
        linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org
Subject: [PATCH V3 0/3] Begin converting kmap calls to kmap_local_page()

From: Ira Weiny <ira.weiny@...el.com>

Changes from V2[1]:
	Update this cover letter
	Update commit messages
	From Matthew Wilcox
		Put functions in highmem.h rather than pagemap.h
	Investigate 0-day build errors.
		AFAICT the patches were applied to the wrong tree and caused
		build errors.

There are many places in the kernel where kmap is used for a simple memory
operation like memcpy, memset, or memmove and then the page is unmapped.

This kmap/mem*/kunmap pattern is mixed between kmap and kmap_atomic uses.  All
of them could use kmap_atomic() which is faster.  However, kmap_atomic() is
being deprecated in favor of kmap_local_page().

Use kmap_local_page() in the existing page operations.  Lift
memcpy_[to|from]_page to highmem.h.  Remove memzero_page() and use zero_user()
instead.  Add memcpy_page(), memmove_page(), and memset_page() to be used in
future patches.  Finally, add BUG_ON()s to check for any miss use of the API
and prevent data corruption in the same way zero_user() does.

This is V3 to get into 5.11 so that we can start to convert all the various
subsystems in 5.12.[2]

These are based on tip/core/mm.  I'm sending to Andrew and Thomas but I'm
expecting these to go through tip/core/mm via Thomas if that is ok with Andrew.

[1] https://lore.kernel.org/lkml/20201207225703.2033611-1-ira.weiny@intel.com/
[2] https://lore.kernel.org/lkml/20201204160504.GH1563847@iweiny-DESK2.sc.intel.com/


Ira Weiny (2):
  mm/highmem: Remove deprecated kmap_atomic
  mm/highmem: Lift memcpy_[to|from]_page to core

 include/linux/highmem.h | 81 ++++++++++++++++++++++++++++++++++-------
 lib/iov_iter.c          | 26 ++-----------
 2 files changed, 70 insertions(+), 37 deletions(-)

-- 
2.28.0.rc0.12.gb6a658bd00c9

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ