Revision e2093926a098a8ccf0f1d10f6df8dad452cb28d3 authored by Ross Zwisler on 02 June 2017, 21:46:37 UTC, committed by Linus Torvalds on 02 June 2017, 22:07:37 UTC
We currently have two related PMD vs PTE races in the DAX code. These can both be easily triggered by having two threads reading and writing simultaneously to the same private mapping, with the key being that private mapping reads can be handled with PMDs but private mapping writes are always handled with PTEs so that we can COW. Here is the first race: CPU 0 CPU 1 (private mapping write) __handle_mm_fault() create_huge_pmd() - FALLBACK handle_pte_fault() passes check for pmd_devmap() (private mapping read) __handle_mm_fault() create_huge_pmd() dax_iomap_pmd_fault() inserts PMD dax_iomap_pte_fault() does a PTE fault, but we already have a DAX PMD installed in our page tables at this spot. Here's the second race: CPU 0 CPU 1 (private mapping read) __handle_mm_fault() passes check for pmd_none() create_huge_pmd() dax_iomap_pmd_fault() inserts PMD (private mapping write) __handle_mm_fault() create_huge_pmd() - FALLBACK (private mapping read) __handle_mm_fault() passes check for pmd_none() create_huge_pmd() handle_pte_fault() dax_iomap_pte_fault() inserts PTE dax_iomap_pmd_fault() inserts PMD, but we already have a PTE at this spot. The core of the issue is that while there is isolation between faults to the same range in the DAX fault handlers via our DAX entry locking, there is no isolation between faults in the code in mm/memory.c. This means for instance that this code in __handle_mm_fault() can run: if (pmd_none(*vmf.pmd) && transparent_hugepage_enabled(vma)) { ret = create_huge_pmd(&vmf); But by the time we actually get to run the fault handler called by create_huge_pmd(), the PMD is no longer pmd_none() because a racing PTE fault has installed a normal PMD here as a parent. This is the cause of the 2nd race. The first race is similar - there is the following check in handle_pte_fault(): } else { /* See comment in pte_alloc_one_map() */ if (pmd_devmap(*vmf->pmd) || pmd_trans_unstable(vmf->pmd)) return 0; So if a pmd_devmap() PMD (a DAX PMD) has been installed at vmf->pmd, we will bail and retry the fault. This is correct, but there is nothing preventing the PMD from being installed after this check but before we actually get to the DAX PTE fault handlers. In my testing these races result in the following types of errors: BUG: Bad rss-counter state mm:ffff8800a817d280 idx:1 val:1 BUG: non-zero nr_ptes on freeing mm: 15 Fix this issue by having the DAX fault handlers verify that it is safe to continue their fault after they have taken an entry lock to block other racing faults. [ross.zwisler@linux.intel.com: improve fix for colliding PMD & PTE entries] Link: http://lkml.kernel.org/r/20170526195932.32178-1-ross.zwisler@linux.intel.com Link: http://lkml.kernel.org/r/20170522215749.23516-2-ross.zwisler@linux.intel.com Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Reported-by: Pawel Lebioda <pawel.lebioda@intel.com> Reviewed-by: Jan Kara <jack@suse.cz> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Matthew Wilcox <mawilcox@microsoft.com> Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com> Cc: Pawel Lebioda <pawel.lebioda@intel.com> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Xiong Zhou <xzhou@redhat.com> Cc: Eryu Guan <eguan@redhat.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent d0f0931
config
#!/bin/bash
# Manipulate options in a .config file from the command line
myname=${0##*/}
# If no prefix forced, use the default CONFIG_
CONFIG_="${CONFIG_-CONFIG_}"
usage() {
cat >&2 <<EOL
Manipulate options in a .config file from the command line.
Usage:
$myname options command ...
commands:
--enable|-e option Enable option
--disable|-d option Disable option
--module|-m option Turn option into a module
--set-str option string
Set option to "string"
--set-val option value
Set option to value
--undefine|-u option Undefine option
--state|-s option Print state of option (n,y,m,undef)
--enable-after|-E beforeopt option
Enable option directly after other option
--disable-after|-D beforeopt option
Disable option directly after other option
--module-after|-M beforeopt option
Turn option into module directly after other option
commands can be repeated multiple times
options:
--file config-file .config file to change (default .config)
--keep-case|-k Keep next symbols' case (dont' upper-case it)
$myname doesn't check the validity of the .config file. This is done at next
make time.
By default, $myname will upper-case the given symbol. Use --keep-case to keep
the case of all following symbols unchanged.
$myname uses 'CONFIG_' as the default symbol prefix. Set the environment
variable CONFIG_ to the prefix to use. Eg.: CONFIG_="FOO_" $myname ...
EOL
exit 1
}
checkarg() {
ARG="$1"
if [ "$ARG" = "" ] ; then
usage
fi
case "$ARG" in
${CONFIG_}*)
ARG="${ARG/${CONFIG_}/}"
;;
esac
if [ "$MUNGE_CASE" = "yes" ] ; then
ARG="`echo $ARG | tr a-z A-Z`"
fi
}
txt_append() {
local anchor="$1"
local insert="$2"
local infile="$3"
local tmpfile="$infile.swp"
# sed append cmd: 'a\' + newline + text + newline
cmd="$(printf "a\\%b$insert" "\n")"
sed -e "/$anchor/$cmd" "$infile" >"$tmpfile"
# replace original file with the edited one
mv "$tmpfile" "$infile"
}
txt_subst() {
local before="$1"
local after="$2"
local infile="$3"
local tmpfile="$infile.swp"
sed -e "s:$before:$after:" "$infile" >"$tmpfile"
# replace original file with the edited one
mv "$tmpfile" "$infile"
}
txt_delete() {
local text="$1"
local infile="$2"
local tmpfile="$infile.swp"
sed -e "/$text/d" "$infile" >"$tmpfile"
# replace original file with the edited one
mv "$tmpfile" "$infile"
}
set_var() {
local name=$1 new=$2 before=$3
name_re="^($name=|# $name is not set)"
before_re="^($before=|# $before is not set)"
if test -n "$before" && grep -Eq "$before_re" "$FN"; then
txt_append "^$before=" "$new" "$FN"
txt_append "^# $before is not set" "$new" "$FN"
elif grep -Eq "$name_re" "$FN"; then
txt_subst "^$name=.*" "$new" "$FN"
txt_subst "^# $name is not set" "$new" "$FN"
else
echo "$new" >>"$FN"
fi
}
undef_var() {
local name=$1
txt_delete "^$name=" "$FN"
txt_delete "^# $name is not set" "$FN"
}
if [ "$1" = "--file" ]; then
FN="$2"
if [ "$FN" = "" ] ; then
usage
fi
shift 2
else
FN=.config
fi
if [ "$1" = "" ] ; then
usage
fi
MUNGE_CASE=yes
while [ "$1" != "" ] ; do
CMD="$1"
shift
case "$CMD" in
--keep-case|-k)
MUNGE_CASE=no
continue
;;
--refresh)
;;
--*-after|-E|-D|-M)
checkarg "$1"
A=$ARG
checkarg "$2"
B=$ARG
shift 2
;;
-*)
checkarg "$1"
shift
;;
esac
case "$CMD" in
--enable|-e)
set_var "${CONFIG_}$ARG" "${CONFIG_}$ARG=y"
;;
--disable|-d)
set_var "${CONFIG_}$ARG" "# ${CONFIG_}$ARG is not set"
;;
--module|-m)
set_var "${CONFIG_}$ARG" "${CONFIG_}$ARG=m"
;;
--set-str)
# sed swallows one level of escaping, so we need double-escaping
set_var "${CONFIG_}$ARG" "${CONFIG_}$ARG=\"${1//\"/\\\\\"}\""
shift
;;
--set-val)
set_var "${CONFIG_}$ARG" "${CONFIG_}$ARG=$1"
shift
;;
--undefine|-u)
undef_var "${CONFIG_}$ARG"
;;
--state|-s)
if grep -q "# ${CONFIG_}$ARG is not set" $FN ; then
echo n
else
V="$(grep "^${CONFIG_}$ARG=" $FN)"
if [ $? != 0 ] ; then
echo undef
else
V="${V/#${CONFIG_}$ARG=/}"
V="${V/#\"/}"
V="${V/%\"/}"
V="${V//\\\"/\"}"
echo "${V}"
fi
fi
;;
--enable-after|-E)
set_var "${CONFIG_}$B" "${CONFIG_}$B=y" "${CONFIG_}$A"
;;
--disable-after|-D)
set_var "${CONFIG_}$B" "# ${CONFIG_}$B is not set" "${CONFIG_}$A"
;;
--module-after|-M)
set_var "${CONFIG_}$B" "${CONFIG_}$B=m" "${CONFIG_}$A"
;;
# undocumented because it ignores --file (fixme)
--refresh)
yes "" | make oldconfig
;;
*)
usage
;;
esac
done
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...