From: Lipeng
Date: Tue, 19 Dec 2017 12:02:24 +0800
> @@ -2651,6 +2651,19 @@ static int hns3_get_ring_config(struct hns3_nic_priv
> *priv)
> return ret;
> }
>
> +static void hns3_put_ring_config(struct hns3_nic_priv *priv)
> +{
> + struct hnae3_handle *h = priv->ae_handle;
> + u1
The current rule for generating the {shipped,extra}-certs.c source files
might create an invalid C source file, containing redefinitions of the
same variables:
CC [M] net/wireless/shipped-certs.o
net/wireless/shipped-certs.c:686:10: error: redefinition of
'shipped_regdb_certs'
const u8 shippe
Hi all,
This patch series updates the Broadcom STB Bus Interface Unit controller to
support newer chips such as 7260, 7268, 7271 and 7278. These chips require
additional tuning in order to provide the expected bus throughput.
In the process, we need to re-organize the common.c file a little bit i
Correct the Device Tree bindings for the HIF_CPUBIUCTRL node whose
compatible string is actually brcm,bcm-cpu-biu-ctrl. Also
document in the binding the fallback property
("brcm,brcmstb-cpu-biu-ctrl") and update the example accordingly.
Signed-off-by: Florian Fainelli
---
.../devicetree/bindings
Broadcom's Brahma-B53 CPU is an ARMv8A processor used on a number of
DSL, Cable Modem and Set-top-box SoCs.
Acked-by: Rob Herring
Signed-off-by: Florian Fainelli
---
Documentation/devicetree/bindings/arm/cpus.txt | 1 +
1 file changed, 1 insertion(+)
diff --git a/Documentation/devicetree/bindi
We may need access to family_id and product_id fairly early on boot for
other parts of the code (e.g: biuctrl.c), so split the initialization
between an early_init() and an arch_initcall() which allows us to do
that.
Signed-off-by: Florian Fainelli
---
drivers/soc/bcm/brcmstb/common.c | 27 +
Being called during early_initcall() is early enough that it occurs
before SMP initialization, which is all we care about for the Bus
Interface Unit configuration.
This solves lack of BIU initialization on ARM64 platforms where we do
not have an anchor where to put the BIU initialization (since th
In preparation for saving/restoring additional registers required on
some newer platforms (7268, 7271, 7278), migrate the code to use enums
and helper functions to access registers.
Signed-off-by: Florian Fainelli
---
drivers/soc/bcm/brcmstb/biuctrl.c | 75 ++-
Add definitions for B53 systems register: CPU_MCP_FLOW_REG and
CPU_WRITEBACK_CTRL_REG. These register will be saved and restored
accordingly.
Signed-off-by: Florian Fainelli
---
drivers/soc/bcm/brcmstb/biuctrl.c | 8 +++-
1 file changed, 7 insertions(+), 1 deletion(-)
diff --git a/drivers/s
In order to achieve expected MCP bus throughput on 3 particular chips:
7268, 7271 and 7278, do the appropriate programming of the MCP
interface: increase number of MCP write credits, turn on write-back
throttling when present.
Signed-off-by: Florian Fainelli
---
drivers/soc/bcm/brcmstb/biuctrl.c
On Wed, Dec 13, 2017 at 03:23:21PM +0530, Viresh Kumar wrote:
> Currently the schedutil governor overwrites the sg_cpu->flags field on
> every call to the utilization handler. It was pretty good as the initial
> implementation of utilization handlers, there are several drawbacks
> though.
>
> The
On Broadcom Brahma-B53 CPUs, the CPU_CREDIT_REG offset got moved to
0x0b0 instead of 0x184, correct this such that we correcty
enable/disable write-pairing for these chips.
Signed-off-by: Florian Fainelli
---
drivers/soc/bcm/brcmstb/biuctrl.c | 24 +---
1 file changed, 21 ins
On 12/19/2017 12:15 PM, Joe Perches wrote:
drivers/char/ipmi/ipmi_msghandler.c| 17 +++---
For ipmi:
Acked-by: Corey Minyard
In preparation for fixing and changing values in the CPU_CREDIT_REG
register for B53-based systems, make the offset parameterized.
Signed-off-by: Florian Fainelli
---
drivers/soc/bcm/brcmstb/biuctrl.c | 11 ++-
1 file changed, 6 insertions(+), 5 deletions(-)
diff --git a/drivers/soc/bcm
Printing kernel addresses should be done in limited circumstances, mostly
for debugging purposes. Printing out the virtual memory layout at every
kernel bootup doesn't really fall into this category so delete the prints.
There are other ways to get the same information.
Signed-off-by: Laura Abbott
On 19 December 2017 at 18:52, Keno Fischer wrote:
> Yes it seems like an EINTR return should be considered a bug, so please drop
> this from your patch queue. Thanks for the follow up.
Okay -- thanks for the info.
Cheers,
Michael
> On Dec 19, 2017 14:57, "Michael Kerrisk (man-pages)"
> wrote
On Tue, Dec 19, 2017 at 12:27:30PM -0600, Eric W. Biederman wrote:
> Dave Jones writes:
>
> > On Mon, Dec 18, 2017 at 03:50:52PM -0800, Linus Torvalds wrote:
> >
> > > But I don't see what would have changed in this area recently.
> > >
> > > Do you end up saving the seeds that cause c
On Tue, Dec 19, 2017 at 09:52:27AM -0800, rao.sho...@oracle.com wrote:
> @@ -129,6 +130,7 @@ int __kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t
> flags, size_t nr,
>
> for (i = 0; i < nr; i++) {
> void *x = p[i] = kmem_cache_alloc(s, flags);
> +
> if (!x) {
On Tue, Dec 19, 2017 at 10:48 AM, Peter Zijlstra wrote:
> On Tue, Dec 19, 2017 at 09:47:12AM -0800, Joel Fernandes wrote:
>> Since the recent remote cpufreq callback work, its possible that a cpufreq
>> update is triggered from a remote CPU. For single policies however, the
>> current
>> code use
On Tue, Dec 12, 2017 at 12:50:40PM +0100, Ingo Molnar wrote:
>
> * Claudio Scordino wrote:
>
> > From: Juri Lelli
>
> > - ~(SCHED_FLAG_RESET_ON_FORK | SCHED_FLAG_RECLAIM))
> > + ~(SCHED_FLAG_RESET_ON_FORK | SCHED_FLAG_RECLAIM |
> > + SCHED_FLAG_DL_OVERRUN)
On Sat, Dec 09, 2017 at 09:03:49AM +0100, Filippo Sironi wrote:
> ... since total = sched_avg_period() + delta can yield 0x1,
> which results in a division by 0, given that div_u64() takes a u32
> divisor. Use div64_u64() instead.
>
> divide error: [#1] SMP
> CPU: 7 PID: 0 Comm: swap
On Tue, 19 Dec 2017 09:52:27 -0800 rao.sho...@oracle.com wrote:
> diff --git a/mm/slab_common.c b/mm/slab_common.c
> index c8cb367..06fd12c 100644
> --- a/mm/slab_common.c
> +++ b/mm/slab_common.c
> @@ -20,6 +20,7 @@
> #include
> #include
> #include
> +#include
>
> #define CREATE_TRACE_
On Tue, 19 Dec 2017, rao.sho...@oracle.com wrote:
> This patch updates kfree_rcu to use new bulk memory free functions as they
> are more efficient. It also moves kfree_call_rcu() out of rcu related code to
> mm/slab_common.c
It would be great to have separate patches so that we can review it
pro
Upload of intial version of hyper_DMABUF driver enabling
DMA_BUF exchange between two different VMs in virtualized
platform based on hypervisor such as KVM or XEN.
Hyper_DMABUF drv's primary role is to import a DMA_BUF
from originator then re-export it to another Linux VM
so that it can be mapped
High-level description of hyper_dmabuf driver has been added
to "Documentation" directory.
Signed-off-by: Dongwon Kim
---
Documentation/hyper-dmabuf-sharing.txt | 734 +
1 file changed, 734 insertions(+)
create mode 100644 Documentation/hyper-dmabuf-sharing.txt
Now we re-use dma_buf instead of exporting it via normal process
(including new mappings). For this, hyper_dmabuf list entries can
be searched with "struct dma_buf*". Also, ioctl (export_remote) is
modified to just return hyper_dmabuf_id if the specific dmabuf
has already been exported to the targe
unexporting on exporter's side now have two options, one is
, that just remove and free everything to literally "disconnect"
from importer, the other is just to return fail if any apps
running on importer is still attached or DMAing. Currently whether
forcing or unforcing it is determined by how "F
Now, released hyper_dmabuf_ids are stored in a stack -
(hyper_dmabuf_private.id_queue) for reuse. This is to prevent
overflow of ids for buffers. We also limit maximum number for
the id to 1000 for the stability and optimal performance.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/Mak
From: Mateusz Polrola
When error occurs while sharing pages, all pages already shared
needs to be un-shared and proper error code has to be returned.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c | 6 ++-
.../xen/hyper_dmabuf/xen/hyper_dmabuf_xen_shm.c|
From: Mateusz Polrola
This patch addresses several synchronization issues while sharing
DMA_BUF with another VM.
1. Set WAIT_AFTER_SYNC_REQ to false by default to prevent possible
performance degradation when waing for the response for every
syncrhonization request to exporter VM.
2. Remo
From: Kai Chen
Make sure hy_drv_priv is freed before exiting in several places
in hyper_dmabuf_drv_init.
v2: unlocking mutex before freeing hy_drv_priv when bknd_ops->init
fails
Signed-off-by: Kai Chen
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c | 7 ++-
1
From: Mateusz Polrola
Introduced additional init and cleanup routines in the backend
API structure that might be useful for hypervisors other than Xen.
Signed-off-by: Mateusz Polrola
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c | 14 ++
driver
From: Mateusz Polrola
hyper_dmabuf driver is generic driver that is designed to work
with any hypervisor with various backend implementations. So
moving out its device node out of /dev/xen.
Signed-off-by: Mateusz Polrola
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv
From: Mateusz Polrola
Changed data type of reference for the group of pages to be shared
unsigned long in case it is direct representation of the memory
address.
Signed-off-by: Mateusz Polrola
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.h | 2 +-
drivers/x
From: Mateusz Polrola
Wrong vmid is used in case of sending HYPER_DMABUF_EXPORT_FD_FAILED
message. Instead of vmid, hypder dmabuf id is being used.
Signed-off-by: Mateusz Polrola
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c | 2 +-
1 file changed, 1 insertion(+
From: Mateusz Polrola
It is needed to freeing already-mapped pages if it gets error
before finishing mapping all pages.
Signed-off-by: Dongwon Kim
---
.../xen/hyper_dmabuf/xen/hyper_dmabuf_xen_shm.c| 43 +++---
1 file changed, 38 insertions(+), 5 deletions(-)
diff --git a/
To make hyper_dmabuf_export_remote_ioctl more compact and readable,
a new function call, 'fastpath_export' is created to replace a routine
in hyper_dmabuf_export_remote_ioctl for the case requested buffer for
exporting is already in the LIST (exported previously).
Signed-off-by: Dongwon Kim
---
Fixing all styling problems caught by checkpatch.pl
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c| 53 ++--
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.h| 6 +-
drivers/xen/hyper_dmabuf/hyper_dmabuf_event.c | 12 +-
drivers/xen/hyper_dmabuf/hyp
To make type's name compact, *_backend_ops is changed to '*_bknd_ops'. Also
'ops' is now changed to 'bknd_ops' to clarify it is a data structure with
entry points of 'backend' operations.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c| 14 +--
drivers/
Added missing mutex_unlock to make sure mutex is unlocked
before returning.
Also, moved spinlock lock/unlock into hyper_dmabuf_send_event
and remove checking on spinlock (with assumption caller does
the spinlock in advance) to make it more straight forward.
This patch includes a couple of minor m
Removed prefix "hyper_dmabuf" from backend functions and static func
(except for driver APIs) and add 'be' after 'xen' in backend function
calls to show those are backend APIs.
Also, modified some of function names for clarification and addressed
some missing errors and warnings in hyper_dmabuf_l
From: Mateusz Polrola
This change adds two query items, 'HYPER_DMABUF_QUERY_PRIV_INFO_SIZE'
and 'HYPER_DMABUF_QUERY_PRIV_INFO', for retrieving buffer's private
info and its size.
'info' is an address of user-space buffer (user application provides
this) ,where private data will be copied in case
From: Mateusz Polrola
Added mutex to export_fd ioctl to prevent double pages mapping of the
same buffer to prevent race condition when two consumers are trying to
map the same buffer on importer VM.
Also locked mutex before sending request via xen communication channel
to prevent req_pending ove
Comm env intialization is now scheduled to be done
when xenstore is initialized. This scheduling is done
in driver's init routine.
Also, adding a recursively scheduled routine
that monitors any new tx ch setup from other domains
and automaticlaly configure rx channel accordingly
(every 10 sec).
O
sgt_info->priv will be used to store user private info passed in
ioctl. Data in sgt_info->priv is transfered via comm channel to
the importer VM whenever DMA_BUF is exported to keep the private
data synchroized across VMs.
This patch also adds hyper_dmabuf_send_export_msg that replaces
some of exp
This patch fixes several defects on event handling
including:
1. Imported sgt info and exported sgt info now have
buffer for private data (priv) with variable size
2. Now user input to export_remote_ioctl contain sz_priv,
which specifies size of private data (e.g. meta data)
3. Increased m
hyper_dmabuf driver on importing domain now generates event
every time new hyper_dmabuf is available (visible) to the
importer.
Each event comes with 128 byte private data, which can
contain any meta data or user data specific to the originator
of DMA BUF.
Signed-off-by: Dongwon Kim
---
drivers
From: Mateusz Polrola
Need to update random keys when reusing hyper_dmabuf_id
in the list to increase security
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_id.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/xen/hyper_dmabuf/hyper_dmabu
From: Mateusz Polrola
Using a word, 'private' is not recommended because of conflict
with language keyword when compiling with C++.
So changing those to 'priv'.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c | 8
include/uapi/xen/hyper_dmabuf.h
Make sure to free buffers before returning to prevent memory leaks
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c | 19 +++-
drivers/xen/hyper_dmabuf/hyper_dmabuf_msg.c| 9 +++-
drivers/xen/hyper_dmabuf/hyper_dmabuf_ops.c| 6 ++-
drivers/x
Re-orginized source code for more intuitive structure
For this,
1. driver's file operations other than ioctls have been moved to
hyper_dmabuf_drv.c.
2. Separated out dma-buf operations from hyper_dmabuf_ops.c
and put those in a new file, 'hyper_dmabuf_ops.c'. Remaining part
(SGT core management)
From: Mateusz Polrola
Extending DMA bitmask of hyper_dmabuf device to cover whole
address space driver may access.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/xen/hyper_dmabuf/hyper_dma
Cleaned up and corrected error codes and condition in various
error check routines. Also added proper err messages when func
returns error.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c| 14 +++--
drivers/xen/hyper_dmabuf/hyper_dmabuf_id.c | 2 +-
dr
The length of hyper_dmabuf_id is increased to 128bit by adding
random key (96bit) to the id. This is to prevent possible leak
of the id by guessing on importer VM (by unauthorized application).
hyper_dmabuf_id_t is now defined as,
typedef struct {
int id;
int rng_k
On 12/19/2017 11:12 AM, Matthew Wilcox wrote:
On Tue, Dec 19, 2017 at 09:52:27AM -0800, rao.sho...@oracle.com wrote:
This patch updates kfree_rcu to use new bulk memory free functions as they
are more efficient. It also moves kfree_call_rcu() out of rcu related code to
mm/slab_common.c
Signed
Reordering clean-up procedure
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c | 37 +--
1 file changed, 24 insertions(+), 13 deletions(-)
diff --git a/drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c
b/drivers/xen/hyper_dmabuf/hyper_dmabuf_ioct
List of queries is re-defined. Now it supports following
items:
enum hyper_dmabuf_query {
DMABUF_QUERY_TYPE = 0x10,
DMABUF_QUERY_EXPORTER,
DMABUF_QUERY_IMPORTER,
DMABUF_QUERY_SIZE,
DMABUF_QUERY_BUSY,
Adding ID check to make sure a dma-buf is exported externally
since hyper_dmabuf only allows to export a dmabuf to a different
VM.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c | 6 ++
1 file changed, 6 insertions(+)
diff --git a/drivers/xen/hyper_dmabuf/hyper
From: Mateusz Polrola
It is required to release all allocated buffers when the application
crashes. With the change, hyper_dmabuf_sgt_info includes file pointers
for the driver. If it's released unexpectedly, the driver is now
unexporting all already-exported buffers to prevent memory leak.
In c
Quoting Jerome Brunet (2017-12-19 00:33:29)
> Nothing really prevents a provider from (trying to) register a clock
> without providing the clock ops structure.
>
> We do check the individual fields before using them, but not the
> structure pointer itself. This may have the usual nasty consequence
All shared pages need to be read-only from importer's
point of view to prevent the buffer from being corrupted.
This patch may need to be reverted if we find a better
way to protect the original content in this sharing
model.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/xen/hyper_dma
From: Mateusz Polrola
Enabled built-in compilation option of hyper_dmabuf driver.
Also, moved backend initialization into open() to remove
its dependencies on Kernel booting sequence.
hyper_dmabuf.h is now installed as one of standard header
files of Kernel.
This patch also addresses possible m
From: Mateusz Polrola
Added error checking on the result of function call,
dma_buf_map_attachment
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c
b/drivers/xen/h
On Fri, Dec 08, 2017 at 12:07:54PM -0800, subhra mazumdar wrote:
> +static inline void
> +sd_context_switch(struct sched_domain *sd, struct rq *rq, int util)
> +{
> + struct sched_group *sg_cpu;
> +
> + /* atomically add/subtract the util */
> + sg_cpu = sd->sg_cpu;
> + if (util >
From: Mateusz Polrola
To prevent overhead when a DMA BUF needs to be exported right after
it is unexported, a marginal delay is introduced in unexporting process.
This adds a probation period to the unexporting process. If the same
DMA_BUF is requested to be exported agagin, unexporting process i
Now, use CONFIG_HYPER_DMABUF_XEN as a configuration option
for building hyper_dmabuf for Xen hypervisor.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c
hyper_dmabuf's sync_request (previously hyper_dmabuf_sync_request_
and_wait) now does not wait for the response from exporter if
WAIT_AFTER_SYNC_REQ==0. This is to prevent peformance degradation
due to the communication latency while doing indirect hyper DMABUF
synchronization.
This patch also inc
Make sure list of mapping/attaching activities on impoter VM is not
empty before doing unmapping/detaching shadow DMA BUF for indirect
synchronization.
Signed-off-by: Dongwon Kim
---
.../xen/hyper_dmabuf/hyper_dmabuf_remote_sync.c| 68 +-
1 file changed, 53 insertions(+),
From: Michał Janiszewski
This adds two entries in SYSFS with information about imported
and exported entries. The information exposed contains details
about number of pages, whether a buffer is valid or not, and
importer/exporter count.
Sysfs for hyper_dmabuf can be enabled by setting a new conf
Redefined nents_last, which means # of gref in the last page
of lvl2 table in any situation even if it is same as REFS_PER_PAGE.
With this, loop can be simplified with less condition check.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/xen/hyper_dmabuf_xen_shm.c | 8 +---
1 file ch
Set the license of the driver to "GPL and MIT-X dual" and owner
to "Intel". Also attached license term to all source and header
files
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_conf.h | 26 ++
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c| 32
From: Mateusz Polrola
Checking whether buffer has more than two pages should be done
by evaluating nents > 1 instead of i > 1 to properly cover the
case when nents == 2.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_imp.c | 2 +-
1 file changed, 1 insertion(+), 1 deletio
From: Mateusz Polrola
Replaces printk to debug macros
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c| 4 +-
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.h| 4 ++
drivers/xen/hyper_dmabuf/hyper_dmabuf_imp.c| 46 +-
drivers/xen/hyper
From: Mateusz Polrola
When exporter or importer is disconnected, ring buffer should be
reinitialzed, otherwise on next reconnection exporter/importer will
receive old requests/responses remaining in the ring buffer, which are
not valid anymore.
This patch also blocks back ring irq until communic
For adoption of hyper_dmabuf driver to various hypervisors
other than Xen, a "backend" layer is defined and separated out
from existing one-body structure.
"Backend" is basically a list of entry points of function calls
that provides method to do Kernel's page-level sharing and inter
VMs communica
From: Mateusz Polrola
In hyper_dmabuf_export_remote, page_info->pages needs to
be freed before freeing page_info.
Also, info_entry in hyper_dmabuf_remove_exported/imported
and hyper_dmabuf_remove_exporter/importer_ring needs to
be freed after removal of an entry.
Signed-off-by: Dongwon Kim
---
From: Mateusz Polrola
Immediate clean up of buffer is not possible if the buffer is
actively used by importer. In this case, we need to postpone
freeing hyper_DMABUF until the last consumer unmaps and releases
the buffer on impoter VM. New reference count is added for tracking
usage by importers.
Now relaese funcs checks f_count for the file instead of
our own refcount because it can't track dma_buf_get.
Also, importer now sends out HYPER_DMABUF_FIRST_EXPORT
to let the exporter know corresponding dma-buf has ever
exported on importer's side. This is to cover the case
where exporter exports
Need a new index, k in hyper_dmabuf_extract_pgs function for
picking up a correct n-th page in contigous memory space.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_imp.c | 11 ++-
1 file changed, 6 insertions(+), 5 deletions(-)
diff --git a/drivers/xen/hyper_dmab
Importer now sends a synchronization request to the
exporter when any of DMA_BUF operations on imported
Hyper_DMABUF is executed (e.g dma_buf_map and dma_buf_unmap).
This results in a creation of shadow DMA_BUF and exactly same
DMA_BUF operation to be executed on it.
The main purpose of this is to
From: Mateusz Polrola
This introduces use of xenstore for creating and managing
communication channels between two VMs in the system.
When hyper_dmabuf driver is loaded in the service VM (host OS),
a new xenstore directory, "/local/domain//data/hyper_dmabuf"
is created in xenstore filesystem. Wh
Use workqueue mechanism to delay message parsing done
after exiting from ISR to reduce ISR execution time.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.c| 13 ++
drivers/xen/hyper_dmabuf/hyper_dmabuf_drv.h| 5 +
drivers/xen/hyper_dmabuf/hyper_dmabuf
From: Mateusz Polrola
If shared pages of buffer were already mapped on importer side, do
not map them again on next request to export fd.
Signed-off-by: Mateusz Polrola
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.c | 20 +++-
1 file changed, 11 i
hyper_dmabuf_importer_ring_setup creates new channel only if
there is no existing downstream communication channel previously
created for the exporter VM.
Signed-off-by: Dongwon Kim
---
drivers/xen/hyper_dmabuf/hyper_dmabuf_imp.c | 13 +++--
drivers/xen/hyper_dmabuf/hyper_dmabuf_ioctl.
On Tue, Dec 19, 2017 at 10:24:52 -0800,
Shaohua Li wrote:
Not sure if this is MD related, but could you please check if this debug patch
changes anything?
I'm doing a build now. I do use md to mirror disk partitions between two disks. I do that on another machine that doesn't exhibit the pro
This looks okay to me.
On Mon, Dec 18, 2017 at 07:26:03PM -0500, Woody Suwalski wrote:
> The 4.15 drm_atomic_helper driver shows a warning during boot (both 32 and
> 64 bit x86)
> It is caused by a mismatch between the result of vmw_enable_vblank() and
> what the drm_atomic_helper expects:
> /.
rpm-lint flagged these as being executable:
kernel-tools.x86_64: W: spurious-executable-perm
/usr/share/man/man8/turbostat.8.gz
kernel-tools.x86_64: W: spurious-executable-perm
/usr/share/man/man8/x86_energy_perf_policy.8.gz
Fix this
Signed-off-by: Laura Abbott
---
tools/power/x86/turbostat
Quoting Dmitry Osipenko (2017-12-18 19:59:06)
> Machine dies if HCLK, SCLK or EMC is disabled, hence mark these clocks
> as critical. Currently some of drivers do not manage clocks properly,
> expecting clocks to be 'always enabled', these clocks are MC and PLL_P
> outputs. Let's mark MC or PLL_P o
On 12/19/2017 11:30 AM, Matthew Wilcox wrote:
On Tue, Dec 19, 2017 at 09:52:27AM -0800, rao.sho...@oracle.com wrote:
@@ -129,6 +130,7 @@ int __kmem_cache_alloc_bulk(struct kmem_cache *s, gfp_t
flags, size_t nr,
for (i = 0; i < nr; i++) {
void *x = p[i] = kmem_cache_allo
rpm lint flagged this:
perf.x86_64: E: executable-sourced-script /etc/bash_completion.d/perf 755
Make this not be executable
Signed-off-by: Laura Abbott
---
tools/perf/Makefile.perf | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefi
On Wed, Dec 13, 2017 at 5:30 AM, Paolo Bonzini wrote:
> The bulk of the MSR bitmap is either immutable, or can be copied from
> the L1 bitmap. By initializing it at VMXON time, and copying the mutable
> parts one long at a time on vmentry (rather than one bit), about 4000
> clock cycles (30%) can
On 12/19/2017 05:41 PM, Joe Perches wrote:
> On Tue, 2017-12-19 at 17:32 +0100, Martin Homuth wrote:
>> This patch fixes various coding style issues in the rtl8712 module as
>> noted by checkpatch.pl related to dereferencing over multiple lines.
> []
>> diff --git a/drivers/staging/rtl8712/rtl8712_
On 12/19/2017 11:33 AM, Christopher Lameter wrote:
On Tue, 19 Dec 2017, rao.sho...@oracle.com wrote:
This patch updates kfree_rcu to use new bulk memory free functions as they
are more efficient. It also moves kfree_call_rcu() out of rcu related code to
mm/slab_common.c
It would be great to
On 12/19/2017 07:09 PM, Wei Wang wrote:
Would you give more details under what circumstances it happened?
Circumstances are unknown as I can't correlate that log line to anything else.
If I have to guess, then it has to do with ipsec events. These tunnel tend to go
up and down a lot and there
From: Jaegeuk Kim
This patch introduces sysfs entries to show the information.
# cat /sys/devices/soc/1da4000.ufshc/health/eol
# cat /sys/devices/soc/1da4000.ufshc/health/length
# cat /sys/devices/soc/1da4000.ufshc/health/lifetimeA
# cat /sys/devices/soc/1da4000.ufshc/health/lifetimeB
# cat
From: Jaegeuk Kim
This patch introduces attribute group to show existing sysfs entries.
Cc: Greg KH
Signed-off-by: Jaegeuk Kim
---
drivers/scsi/ufs/ufshcd.c | 48 +++
drivers/scsi/ufs/ufshcd.h | 2 --
2 files changed, 19 insertions(+), 31 deletions
From: Markus Elfring
Date: Tue, 19 Dec 2017 20:56:56 +0100
Omit an extra message for a memory allocation failure in this function.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring
---
drivers/platform/x86/mlxcpld-hotplug.c | 5 +
1 file changed, 1 in
Hi all,
Commit
44d5c715330c ("x86/mm: Create asm/invpcid.h")
is missing a Signed-off-by from its committer.
--
Cheers,
Stephen Rothwell
On 12/19/2017 1:58 PM, Peter Zijlstra wrote:
On Mon, Dec 18, 2017 at 03:34:49AM -0800, kan.li...@linux.intel.com wrote:
arch/x86/events/core.c | 14 ++
arch/x86/events/intel/ds.c | 8 +++-
2 files changed, 21 insertions(+), 1 deletion(-)
diff --git a/arch/x86/events/c
Hi,
On 19/12/2017 at 20:27:02 +0530, PrasannaKumar Muralidharan wrote:
> Given the fact that setup code is very small and most of it is generic
> code I strongly believe that it is plausible to make use of generic
> code completely. Please have a look at [1] and [2].
>
> 1. https://patchwork.kern
801 - 900 of 1180 matches
Mail list logo