Starting with version 4.2.a, the AXI DMAC controller can report partial
transfers that have been issued.
This change implements computing DMA residue information for transfers,
based on that reported information.
The way this is done, is to dequeue the partial transfers from the FIFO of
partial transfers, store the partial length to the correct segment &
descriptor, and compute the residue before submitting the DMA cookie to the
DMA framework.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
This allows each virtual channel to store information about each transfer
that completed, i.e. which transfer succeeded (or which failed) and if
there was any residue data on each (completed) transfer.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
When calling debugfs functions, there is no need to ever check the
return value. The function can work or not, but the code logic should
never do something different based on this.
Also, because there is no need to save the file dentry, remove the
variable that was saving it as it was never even being used once set.
Cc: Vinod Koul <vkoul@kernel.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: dmaengine@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
tp->coalesce_info is used in rtl_coalesce_info() only, so we can
remove this member. In addition replace phy_ethtool_get_link_ksettings
with a direct access to tp->phydev->speed.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case of a timeout currently ~0 is returned. Callers often just check
whether a certain bit is set and therefore may behave incorrectly.
So let's return -ETIMEDOUT in case of a timeout.
r8168_phy_ocp_read is used in r8168g_mdio_read only, therefore we can
apply the same change.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The packing facility is needed to decode Ethernet meta frames containing
source port and RX timestamping information.
The DSA driver selects CONFIG_PACKING, but the tagger did not, and since
taggers can be now compiled as modules independently from the drivers
themselves, this is an issue now, as CONFIG_PACKING is disabled by
default on all architectures.
Fixes: e53e18a6fe ("net: dsa: sja1105: Receive and decode meta frames")
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Signed-off-by: David S. Miller <davem@davemloft.net>
We were accessing the pointer returned from platform_get_resource before
checking if it was valid, causing an oops if it was not. Move this access
after the call to devm_ioremap_resource which does the validity check.
Signed-off-by: Robert Hancock <hancock@sedsystems.ca>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Julian Wiedmann says:
====================
s390/qeth: updates 2019-06-11
please apply the following patch series for qeth to net-next.
This brings all sorts of cleanups and minor improvements,
primarily for the control IO path.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
We statically allocate 8 cmd buffers on the read channel, when the only
IO left that's still using them is the long-running READ.
Replace this with a single allocated cmd, that gets restarted whenever
the READ completed.
This introduces refcounting for allocated cmds, so that the READ cmd can
survive the IO completion.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current IDX sequence first sends one WRITE cmd to activate the
device, and then sends a second cmd that READs the response.
Using qeth_alloc_cmd(), we can combine this into a single IO with two
command-chained CCWs.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RCD code is the last remaining IO path that doesn't use the
qeth_send_control_data() infrastructure. Doing so allows us to remove
all sorts of custom state machinery and logic in the IRQ handler.
Instead of introducing statically allocated cmd buffers for this single
IO on the data channel, use the new qeth_alloc_cmd() helper.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
qeth currently uses a fixed set of statically allocated cmd buffers for
the read and write IO channels. This (1) doesn't play well with the single
RCD cmd we need to issue on the data channel, (2) doesn't provide the
necessary flexibility for certain IDX improvements, and (3) is also rather
wasteful since the buffers are idle most of the time.
Add a new type of cmd buffer that is dynamically allocated, and keeps
its ccw chain in the DMA data area. Since this touches most callers of
qeth_setup_ccw(), also add a new CCW flags parameter for future usage.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Each cmd buffer maintains a pointer to the IO channel that it was/will
be issued on. So when dealing with cmd buffers, we don't need to pass
around a separate channel pointer.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The vast majority of SETUP-classified trace entries can be moved to
their device-specific trace file. This reduces pollution of the global
SETUP file, and provides a consistent trace view of all activity on the
device.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
OSN currently provides a custom code path to submit IPA cmds, without
waiting for the cmd response. Replace it with qeth_send_ipa_cmd(), which
uses the common qeth_send_control_data() IO infrastructure.
By setting a custom iob->callback, we can now provide feedback to the
caller about whether the cmd has been successfully submitted to HW.
Since the callback then immediately wakes up the reply-waiter object, we
maintain the old behaviour of returning early without waiting for the
response.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The basic MPC initialization sequence is strictly sequential, and
waiting for an available cmd buffer should never be necessary.
So this change only affects the OSN path, where dangling waiters on an
unbounded wait_event() are not desirable. Switch to qeth_get_buffers(),
and let OSN callers deal with -ENOMEM.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When called from qeth_core_probe_device(), qeth_determine_capabilities()
initializes the device's BLKT defaults. From all other callers, the
ccw_device has already been set online and the BLKT setting is skipped.
Clean this up by extracting the BLKT setting into a separate helper that
gets called from the right place.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The completion of a pending READ cmd is processed via
qeth_issue_next_read_cb(). Let this callback also start the next READ
cmd, instead of hardcoding that step into the IRQ handler.
While at it remove the check of the channel state,
__qeth_issue_next_read() already does this.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the tear down sequence in qeth_l?_stop_card() has finished, the
card is guaranteed to be in DOWN state and we don't have to check for
it again.
With this insight we can also remove the redundant setting of
card->state in qeth_l?_set_online()'s error path.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Slightly reduce the complexity of the core xmit path, by replacing some
open-coded logic with the corresponding helpers.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current code suppresses debug entries when an TX buffer completes in
ERROR state with no error indication set in SBALF15.
This was introduced back with
commit 58490f1807 ("qeth: HiperSockets SIGA retry support on CC=2.").
But qeth no longer retries after CC=2, and this sort of suppression
make no sense anymore. Remove it.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During post-migration device tree updates, we can oops in
pseries_update_drconf_memory() if the source device tree has an
ibm,dynamic-memory-v2 property and the destination has a
ibm,dynamic_memory (v1) property. The notifier processes an "update"
for the ibm,dynamic-memory property but it's really an add in this
scenario. So make sure the old property object is there before
dereferencing it.
Fixes: 2b31e3aec1 ("powerpc/drmem: Add support for ibm, dynamic-memory-v2 property")
Cc: stable@vger.kernel.org # v4.16+
Signed-off-by: Nathan Lynch <nathanl@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
While developing KASAN for 64-bit book3s, I hit the following stack
over-read.
It occurs because the hypercall to put characters onto the terminal
takes 2 longs (128 bits/16 bytes) of characters at a time, and so
hvc_put_chars() would unconditionally copy 16 bytes from the argument
buffer, regardless of supplied length. However, udbg_hvc_putc() can
call hvc_put_chars() with a single-byte buffer, leading to the error.
==================================================================
BUG: KASAN: stack-out-of-bounds in hvc_put_chars+0xdc/0x110
Read of size 8 at addr c0000000023e7a90 by task swapper/0
CPU: 0 PID: 0 Comm: swapper Not tainted 5.2.0-rc2-next-20190528-02824-g048a6ab4835b #113
Call Trace:
dump_stack+0x104/0x154 (unreliable)
print_address_description+0xa0/0x30c
__kasan_report+0x20c/0x224
kasan_report+0x18/0x30
__asan_report_load8_noabort+0x24/0x40
hvc_put_chars+0xdc/0x110
hvterm_raw_put_chars+0x9c/0x110
udbg_hvc_putc+0x154/0x200
udbg_write+0xf0/0x240
console_unlock+0x868/0xd30
register_console+0x970/0xe90
register_early_udbg_console+0xf8/0x114
setup_arch+0x108/0x790
start_kernel+0x104/0x784
start_here_common+0x1c/0x534
Memory state around the buggy address:
c0000000023e7980: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
c0000000023e7a00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 f1 f1
>c0000000023e7a80: f1 f1 01 f2 f2 f2 00 00 00 00 00 00 00 00 00 00
^
c0000000023e7b00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
c0000000023e7b80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
==================================================================
Document that a 16-byte buffer is requred, and provide it in udbg.
Signed-off-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Linux kernel tolerates C++ style comments these days. Actually, the
SPDX License tags for .c files start with //.
On the other hand, uapi headers are written in more strict C, where
the C++ comment style is forbidden.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Frederic Barrat <fbarrat@linux.ibm.com>
Acked-by: Andrew Donnellan <ajd@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Ido Schimmel says:
====================
mlxsw: Add support for physical hardware clock
Shalom says:
This patchset adds support for physical hardware clock for Spectrum-1
ASIC only.
Patches #1, #2 and #3 add the ability to query the free running clock
PCI address.
Patches #4 and #5 add two new register, the Management UTC Register and
the Management Pulse Per Second Register.
Patch #6 publishes scaled_ppm_to_ppb() to allow drivers to use it.
Patch #7 adds the physical hardware clock operations.
Patch #8 initializes the physical hardware clock.
Patch #9 adds a selftest for testing the PTP physical hardware clock.
v2 (Richard):
* s/ptp_clock_scaled_ppm_to_ppb/scaled_ppm_to_ppb/
* imply PTP_1588_CLOCK in mlxsw Kconfig
* s/mlxsw_sp1_ptp_update_phc_settime/mlxsw_sp1_ptp_phc_settime/
* s/mlxsw_sp1_ptp_update_phc_adjfreq/mlxsw_sp1_ptp_phc_adjfreq/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Test the PTP Physical Hardware Clock functionality using the "phc_ctl" (a
part of "linuxptp").
The test contains three sub-tests:
* "settime" test
* "adjtime" test
* "adjfreq" test
"settime" test:
* set the PHC time to 0 seconds.
* wait for 120.5 seconds.
* check if PHC time equal to 120.XX seconds.
"adjtime" test:
* set the PHC time to 0 seconds.
* adjust the time by 10 seconds.
* check if PHC time equal to 10.XX seconds.
"adjfreq" test:
* adjust the PHC frequency to be 1% faster.
* set the PHC time to 0 seconds.
* wait for 100.5 seconds.
* check if PHC time equal to 101.XX seconds.
Usage:
$ ./phc.sh /dev/ptp<X>
It is possible to run a subset of the tests, for example:
* To run only the "settime" test:
$ TESTS="settime" ./phc.sh /dev/ptp<X>
Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Reviewed-by: Petr Machata <petrm@mellanox.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Tested-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Extended fw TDC tests with use cases where actions are pre-created and
attached to a filter by reference, i.e. by action index.
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Logan noticed that devm_memremap_pages_release() kills the percpu_ref
drops all the page references that were acquired at init and then
immediately proceeds to unplug, arch_remove_memory(), the backing pages
for the pagemap. If for some reason device shutdown actually collides
with a busy / elevated-ref-count page then arch_remove_memory() should
be deferred until after that reference is dropped.
As it stands the "wait for last page ref drop" happens *after*
devm_memremap_pages_release() returns, which is obviously too late and
can lead to crashes.
Fix this situation by assigning the responsibility to wait for the
percpu_ref to go idle to devm_memremap_pages() with a new ->cleanup()
callback. Implement the new cleanup callback for all
devm_memremap_pages() users: pmem, devdax, hmm, and p2pdma.
Link: http://lkml.kernel.org/r/155727339156.292046.5432007428235387859.stgit@dwillia2-desk3.amr.corp.intel.com
Fixes: 41e94a8513 ("add devm_memremap_pages")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reported-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The p2pdma facility enables a provider to publish a pool of dma
addresses for a consumer to allocate. A genpool is used internally by
p2pdma to collect dma resources, 'chunks', to be handed out to
consumers. Whenever a consumer allocates a resource it needs to pin the
'struct dev_pagemap' instance that backs the chunk selected by
pci_alloc_p2pmem().
Currently that reference is taken globally on the entire provider
device. That sets up a lifetime mismatch whereby the p2pdma core needs
to maintain hacks to make sure the percpu_ref is not released twice.
This lifetime mismatch also stands in the way of a fix to
devm_memremap_pages() whereby devm_memremap_pages_release() must wait for
the percpu_ref ->release() callback to complete before it can proceed to
teardown pages.
So, towards fixing this situation, introduce the ability to store a 'chunk
owner' at gen_pool_add() time, and a facility to retrieve the owner at
gen_pool_{alloc,free}() time. For p2pdma this will be used to store and
recall individual dev_pagemap reference counter instances per-chunk.
Link: http://lkml.kernel.org/r/155727338118.292046.13407378933221579644.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Patch series "mm/devm_memremap_pages: Fix page release race", v2.
Logan audited the devm_memremap_pages() shutdown path and noticed that
it was possible to proceed to arch_remove_memory() before all potential
page references have been reaped.
Introduce a new ->cleanup() callback to do the work of waiting for any
straggling page references and then perform the percpu_ref_exit() in
devm_memremap_pages_release() context.
For p2pdma this involves some deeper reworks to reference count
resources on a per-instance basis rather than a per pci-device basis. A
modified genalloc api is introduced to convey a driver-private pointer
through gen_pool_{alloc,free}() interfaces. Also, a
devm_memunmap_pages() api is introduced since p2pdma does not
auto-release resources on a setup failure.
The dax and pmem changes pass the nvdimm unit tests, and the p2pdma
changes should now pass testing with the pci_p2pdma_release() fix.
Jrme, how does this look for HMM?
This patch (of 6):
The devm_add_action() facility allows a resource allocation routine to
add custom devm semantics. One such user is devm_memremap_pages().
There is now a need to manually trigger
devm_memremap_pages_release(). Introduce devm_release_action() so the
release action can be triggered via a new devm_memunmap_pages() api in a
follow-on change.
Link: http://lkml.kernel.org/r/155727336530.292046.2926860263201336366.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There was the below bug report from Wu Fangsuo.
On the CMA allocation path, isolate_migratepages_range() could isolate
unevictable LRU pages and reclaim_clean_page_from_list() can try to
reclaim them if they are clean file-backed pages.
page:ffffffbf02f33b40 count:86 mapcount:84 mapping:ffffffc08fa7a810 index:0x24
flags: 0x19040c(referenced|uptodate|arch_1|mappedtodisk|unevictable|mlocked)
raw: 000000000019040c ffffffc08fa7a810 0000000000000024 0000005600000053
raw: ffffffc009b05b20 ffffffc009b05b20 0000000000000000 ffffffc09bf3ee80
page dumped because: VM_BUG_ON_PAGE(PageLRU(page) || PageUnevictable(page))
page->mem_cgroup:ffffffc09bf3ee80
------------[ cut here ]------------
kernel BUG at /home/build/farmland/adroid9.0/kernel/linux/mm/vmscan.c:1350!
Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
Modules linked in:
CPU: 0 PID: 7125 Comm: syz-executor Tainted: G S 4.14.81 #3
Hardware name: ASR AQUILAC EVB (DT)
task: ffffffc00a54cd00 task.stack: ffffffc009b00000
PC is at shrink_page_list+0x1998/0x3240
LR is at shrink_page_list+0x1998/0x3240
pc : [<ffffff90083a2158>] lr : [<ffffff90083a2158>] pstate: 60400045
sp : ffffffc009b05940
..
shrink_page_list+0x1998/0x3240
reclaim_clean_pages_from_list+0x3c0/0x4f0
alloc_contig_range+0x3bc/0x650
cma_alloc+0x214/0x668
ion_cma_allocate+0x98/0x1d8
ion_alloc+0x200/0x7e0
ion_ioctl+0x18c/0x378
do_vfs_ioctl+0x17c/0x1780
SyS_ioctl+0xac/0xc0
Wu found it's due to commit ad6b67041a ("mm: remove SWAP_MLOCK in
ttu"). Before that, unevictable pages go to cull_mlocked so that we
can't reach the VM_BUG_ON_PAGE line.
To fix the issue, this patch filters out unevictable LRU pages from the
reclaim_clean_pages_from_list in CMA.
Link: http://lkml.kernel.org/r/20190524071114.74202-1-minchan@kernel.org
Fixes: ad6b67041a ("mm: remove SWAP_MLOCK in ttu")
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reported-by: Wu Fangsuo <fangsuowu@asrmicro.com>
Debugged-by: Wu Fangsuo <fangsuowu@asrmicro.com>
Tested-by: Wu Fangsuo <fangsuowu@asrmicro.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Pankaj Suryawanshi <pankaj.suryawanshi@einfochips.com>
Cc: <stable@vger.kernel.org> [4.12+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When fixing the race conditions between the coredump and the mmap_sem
holders outside the context of the process, we focused on
mmget_not_zero()/get_task_mm() callers in 04f5866e41 ("coredump: fix
race condition between mmget_not_zero()/get_task_mm() and core
dumping"), but those aren't the only cases where the mmap_sem can be
taken outside of the context of the process as Michal Hocko noticed
while backporting that commit to older -stable kernels.
If mmgrab() is called in the context of the process, but then the
mm_count reference is transferred outside the context of the process,
that can also be a problem if the mmap_sem has to be taken for writing
through that mm_count reference.
khugepaged registration calls mmgrab() in the context of the process,
but the mmap_sem for writing is taken later in the context of the
khugepaged kernel thread.
collapse_huge_page() after taking the mmap_sem for writing doesn't
modify any vma, so it's not obvious that it could cause a problem to the
coredump, but it happens to modify the pmd in a way that breaks an
invariant that pmd_trans_huge_lock() relies upon. collapse_huge_page()
needs the mmap_sem for writing just to block concurrent page faults that
call pmd_trans_huge_lock().
Specifically the invariant that "!pmd_trans_huge()" cannot become a
"pmd_trans_huge()" doesn't hold while collapse_huge_page() runs.
The coredump will call __get_user_pages() without mmap_sem for reading,
which eventually can invoke a lockless page fault which will need a
functional pmd_trans_huge_lock().
So collapse_huge_page() needs to use mmget_still_valid() to check it's
not running concurrently with the coredump... as long as the coredump
can invoke page faults without holding the mmap_sem for reading.
This has "Fixes: khugepaged" to facilitate backporting, but in my view
it's more a bug in the coredump code that will eventually have to be
rewritten to stop invoking page faults without the mmap_sem for reading.
So the long term plan is still to drop all mmget_still_valid().
Link: http://lkml.kernel.org/r/20190607161558.32104-1-aarcange@redhat.com
Fixes: ba76149f47 ("thp: khugepaged")
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: Michal Hocko <mhocko@suse.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jann Horn <jannh@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A few new fields were added to mmu_gather to make TLB flush smarter for
huge page by telling what level of page table is changed.
__tlb_reset_range() is used to reset all these page table state to
unchanged, which is called by TLB flush for parallel mapping changes for
the same range under non-exclusive lock (i.e. read mmap_sem).
Before commit dd2283f260 ("mm: mmap: zap pages with read mmap_sem in
munmap"), the syscalls (e.g. MADV_DONTNEED, MADV_FREE) which may update
PTEs in parallel don't remove page tables. But, the forementioned
commit may do munmap() under read mmap_sem and free page tables. This
may result in program hang on aarch64 reported by Jan Stancek. The
problem could be reproduced by his test program with slightly modified
below.
---8<---
static int map_size = 4096;
static int num_iter = 500;
static long threads_total;
static void *distant_area;
void *map_write_unmap(void *ptr)
{
int *fd = ptr;
unsigned char *map_address;
int i, j = 0;
for (i = 0; i < num_iter; i++) {
map_address = mmap(distant_area, (size_t) map_size, PROT_WRITE | PROT_READ,
MAP_SHARED | MAP_ANONYMOUS, -1, 0);
if (map_address == MAP_FAILED) {
perror("mmap");
exit(1);
}
for (j = 0; j < map_size; j++)
map_address[j] = 'b';
if (munmap(map_address, map_size) == -1) {
perror("munmap");
exit(1);
}
}
return NULL;
}
void *dummy(void *ptr)
{
return NULL;
}
int main(void)
{
pthread_t thid[2];
/* hint for mmap in map_write_unmap() */
distant_area = mmap(0, DISTANT_MMAP_SIZE, PROT_WRITE | PROT_READ,
MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
munmap(distant_area, (size_t)DISTANT_MMAP_SIZE);
distant_area += DISTANT_MMAP_SIZE / 2;
while (1) {
pthread_create(&thid[0], NULL, map_write_unmap, NULL);
pthread_create(&thid[1], NULL, dummy, NULL);
pthread_join(thid[0], NULL);
pthread_join(thid[1], NULL);
}
}
---8<---
The program may bring in parallel execution like below:
t1 t2
munmap(map_address)
downgrade_write(&mm->mmap_sem);
unmap_region()
tlb_gather_mmu()
inc_tlb_flush_pending(tlb->mm);
free_pgtables()
tlb->freed_tables = 1
tlb->cleared_pmds = 1
pthread_exit()
madvise(thread_stack, 8M, MADV_DONTNEED)
zap_page_range()
tlb_gather_mmu()
inc_tlb_flush_pending(tlb->mm);
tlb_finish_mmu()
if (mm_tlb_flush_nested(tlb->mm))
__tlb_reset_range()
__tlb_reset_range() would reset freed_tables and cleared_* bits, but this
may cause inconsistency for munmap() which do free page tables. Then it
may result in some architectures, e.g. aarch64, may not flush TLB
completely as expected to have stale TLB entries remained.
Use fullmm flush since it yields much better performance on aarch64 and
non-fullmm doesn't yields significant difference on x86.
The original proposed fix came from Jan Stancek who mainly debugged this
issue, I just wrapped up everything together.
Jan's testing results:
v5.2-rc2-24-gbec7550cca10
--------------------------
mean stddev
real 37.382 2.780
user 1.420 0.078
sys 54.658 1.855
v5.2-rc2-24-gbec7550cca10 + "mm: mmu_gather: remove __tlb_reset_range() for force flush"
---------------------------------------------------------------------------------------_
mean stddev
real 37.119 2.105
user 1.548 0.087
sys 55.698 1.357
[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/1558322252-113575-1-git-send-email-yang.shi@linux.alibaba.com
Fixes: dd2283f260 ("mm: mmap: zap pages with read mmap_sem in munmap")
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Jan Stancek <jstancek@redhat.com>
Reported-by: Jan Stancek <jstancek@redhat.com>
Tested-by: Jan Stancek <jstancek@redhat.com>
Suggested-by: Will Deacon <will.deacon@arm.com>
Tested-by: Will Deacon <will.deacon@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Nick Piggin <npiggin@gmail.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: <stable@vger.kernel.org> [4.20+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
