On Mon, Mar 27, 2023 at 8:15 PM Arnd Bergmann wrote:
>
> From: Arnd Bergmann
>
> For a DMA_BIDIRECTIONAL transfer, the caches have to be cleaned
> first to let the device see data written by the CPU, and invalidated
> after the transfer to let the CPU see data written by the device.
>
> riscv
On 03 May 2023 at 08:15 pm, Rob Herring wrote:
On Wed, May 3, 2023 at 12:40 PM Christian Zigotzky
wrote:
On 3. May 2023, at 18:51, Rob Herring wrote:
On Wed, May 3, 2023 at 11:27 AM Christophe Leroy
wrote:
+Rob as he's the commit's Author.
Le 03/05/2023 à 17:46, Christian Zigotzky a
Since commit 1ba3cbf3ec3b ("mm: kfence: improve the performance of
__kfence_alloc() and __kfence_free()"), kfence reports failures in
random places at boot on big endian machines.
The problem is that the new KFENCE_CANARY_PATTERN_U64 encodes the
address of each byte in its value, so it needs to
On Fri May 5, 2023 at 8:13 AM AEST, Douglas Anderson wrote:
> Do a search and replace of:
> - NMI_WATCHDOG_ENABLED => HARD_WATCHDOG_ENABLED
> - watchdog_nmi_ => watchdog_hardlockup_
These are just making prefixes inconsistent again.
If you really want to do a prefix, I would call it hardlockup
On Fri May 5, 2023 at 8:13 AM AEST, Douglas Anderson wrote:
> These are tiny style changes:
> - Add a blank line before a "return".
> - Renames two globals to use the "watchdog_hld" prefix.
Particularly static ones don't really need the namespace prefixes.
Not sure if processed is better than
On Fri May 5, 2023 at 8:13 AM AEST, Douglas Anderson wrote:
> The perf hardlockup detector works by looking at interrupt counts and
> seeing if they change from run to run. The interrupt counts are
> managed by the common watchdog code via its watchdog_timer_fn().
>
> Currently the API between the
On Fri May 5, 2023 at 8:13 AM AEST, Douglas Anderson wrote:
> The code currently in "watchdog_hld.c" is for detecting hardlockups
> using perf, as evidenced by the line in the Makefile that only
> compiles this file if CONFIG_HARDLOCKUP_DETECTOR_PERF is
> defined. Rename the file to prepare for
On Fri May 5, 2023 at 8:13 AM AEST, Douglas Anderson wrote:
> In preparation for the buddy hardlockup detector, rename
> touch_nmi_watchdog() to touch_hardlockup_watchdog() to make it clear
> that it will touch whatever hardlockup detector is configured. We'll
> add a #define for the old name
On Fri May 5, 2023 at 8:13 AM AEST, Douglas Anderson wrote:
> From: Lecopzer Chen
>
> Nobody cares about the return value of watchdog_nmi_enable(),
> changing its prototype to void.
>
Acked-by: Nicholas Piggin
> Signed-off-by: Pingfan Liu
> Signed-off-by: Lecopzer Chen
> Reviewed-by: Petr
On Fri May 5, 2023 at 8:13 AM AEST, Douglas Anderson wrote:
> The real watchdog_update_hrtimer_threshold() is defined in
> watchdog_hardlockup_perf.c. That file is included if
In kernel/watchdog_hld.c.
> CONFIG_HARDLOCKUP_DETECTOR_PERF and the function is defined in that
> file if
On Fri May 5, 2023 at 8:13 AM AEST, Douglas Anderson wrote:
> From: Colin Cross
>
> Implement a hardlockup detector that doesn't doesn't need any extra
> arch-specific support code to detect lockups. Instead of using
> something arch-specific we will use the buddy system, where each CPU
> watches
Gaurav Batra writes:
> Hello Michael,
>
> I agree with your concerns regarding a device been able to access memory
> that doesn't belong to it. That exposure we have today with 2MB TCEs.
> With 2MB TCEs, DMA window size will be big enough, for dedicated
> adapters, that whole memory is going
On Thu, Apr 06, 2023 at 01:21:09AM +0100, Maciej W. Rozycki wrote:
> Rename LINK_RETRAIN_TIMEOUT to PCIE_LINK_RETRAIN_TIMEOUT and make it
> available via "pci.h" for PCI drivers to use.
> +#define PCIE_LINK_RETRAIN_TIMEOUT HZ
This is basically just a rename and move, but since we're touching it
On Thu, Apr 06, 2023 at 01:21:31AM +0100, Maciej W. Rozycki wrote:
> Attempt to handle cases such as with a downstream port of the ASMedia
> ASM2824 PCIe switch where link training never completes and the link
> continues switching between speeds indefinitely with the data link layer
> never
With the recent feature added to enable perf events to use pseudo NMIs
as interrupts on platforms which support GICv3 or later, its now been
possible to enable hard lockup detector (or NMI watchdog) on arm64
platforms. So enable corresponding support.
One thing to note here is that normally
From: Lecopzer Chen
Set safe maximum CPU frequency to 5 GHz in case a particular platform
doesn't implement cpufreq driver.
Although, architecture doesn't put any restrictions on
maximum frequency but 5 GHz seems to be safe maximum given the available
Arm CPUs in the market which are clocked
From: Lecopzer Chen
When lockup_detector_init()->watchdog_hardlockup_probe(), PMU may be
not ready yet. E.g. on arm64, PMU is not ready until
device_initcall(armv8_pmu_driver_init). And it is deeply integrated
with the driver model and cpuhp. Hence it is hard to push this
initialization before
On arm64, NMI support needs to be detected at runtime. Add a weak
function to the perf hardlockup detector so that an architecture can
implement it to detect whether NMIs are available.
Signed-off-by: Douglas Anderson
---
While I won't object to this patch landing, I consider it part of the
From: Colin Cross
Implement a hardlockup detector that doesn't doesn't need any extra
arch-specific support code to detect lockups. Instead of using
something arch-specific we will use the buddy system, where each CPU
watches out for another one. Specifically, each CPU will use its
softlockup
The fact that there watchdog_hardlockup_enable(),
watchdog_hardlockup_disable(), and watchdog_hardlockup_probe() are
declared __weak means that the configured hardlockup detector can
define non-weak versions of those functions if it needs to. Instead of
doing this, the perf hardlockup detector
Do a search and replace of:
- NMI_WATCHDOG_ENABLED => HARD_WATCHDOG_ENABLED
- watchdog_nmi_ => watchdog_hardlockup_
Then update a few comments near where names were changed.
This is specifically to make it less confusing when we want to
introduce the buddy hardlockup detector, which isn't using
In preparation for the buddy hardlockup detector, which wants the same
petting logic as the current perf hardlockup detector, move the code
to watchdog.c. While doing this, rename the global variable to match
others nearby. The arch_touch_nmi_watchdog() function is not renamed
since that is
In preparation for the buddy hardlockup detector where the CPU
checking for lockup might not be the currently running CPU, add a
"cpu" parameter to watchdog_hardlockup_check().
Signed-off-by: Douglas Anderson
---
Changes in v4:
- ("Add a "cpu" param to watchdog_hardlockup_check()") new for v4.
These are tiny style changes:
- Add a blank line before a "return".
- Renames two globals to use the "watchdog_hld" prefix.
- Store processor id in "unsigned int" rather than "int".
- Minor comment rewording.
- Use "else" rather than extra returns since it seemed more symmetric.
Signed-off-by:
The perf hardlockup detector works by looking at interrupt counts and
seeing if they change from run to run. The interrupt counts are
managed by the common watchdog code via its watchdog_timer_fn().
Currently the API between the perf detector and the common code is a
function: is_hardlockup().
The code currently in "watchdog_hld.c" is for detecting hardlockups
using perf, as evidenced by the line in the Makefile that only
compiles this file if CONFIG_HARDLOCKUP_DETECTOR_PERF is
defined. Rename the file to prepare for the buddy hardlockup detector,
which doesn't use perf.
It could be
In preparation for the buddy hardlockup detector, rename
touch_nmi_watchdog() to touch_hardlockup_watchdog() to make it clear
that it will touch whatever hardlockup detector is configured. We'll
add a #define for the old name (touch_nmi_watchdog) so that we don't
have to touch every piece of code
From: Pingfan Liu
hardlockup_detector_event_create() should create perf_event on the
current CPU. Preemption could not get disabled because
perf_event_create_kernel_counter() allocates memory. Instead,
the CPU locality is achieved by processing the code in a per-CPU
bound kthread.
Add a check
From: Lecopzer Chen
Nobody cares about the return value of watchdog_nmi_enable(),
changing its prototype to void.
Signed-off-by: Pingfan Liu
Signed-off-by: Lecopzer Chen
Reviewed-by: Petr Mladek
Signed-off-by: Douglas Anderson
---
I yanked this patch from the mailing lists [1] into my
From: Lecopzer Chen
No reference to WATCHDOG_DEFAULT, remove it.
Signed-off-by: Pingfan Liu
Signed-off-by: Lecopzer Chen
Reviewed-by: Petr Mladek
Signed-off-by: Douglas Anderson
---
I yanked this patch from the mailing lists [1] into my series just to
make it easier to avoid conflicts
The real watchdog_update_hrtimer_threshold() is defined in
watchdog_hardlockup_perf.c. That file is included if
CONFIG_HARDLOCKUP_DETECTOR_PERF and the function is defined in that
file if CONFIG_HARDLOCKUP_CHECK_TIMESTAMP.
The dummy version of the function in "nmi.h" didn't get that quite
right.
This patch series adds the "buddy" hardlockup detector. In brief, the
buddy hardlockup detector can detect hardlockups without arch-level
support by having CPUs checkup on a "buddy" CPU periodically. All the
details are in the patch ("watchdog/hardlockup: detect hard lockups
using secondary
Hello Michael,
I agree with your concerns regarding a device been able to access memory
that doesn't belong to it. That exposure we have today with 2MB TCEs.
With 2MB TCEs, DMA window size will be big enough, for dedicated
adapters, that whole memory is going to be mapped "direct". Which
When DMA window is backed by 2MB TCEs, the DMA address for the mapped
page should be the offset of the page relative to the 2MB TCE. The code
was incorrectly setting the DMA address to the beginning of the TCE
range.
Mellanox driver is reporting timeout trying to ENABLE_HCA for an SR-IOV
ethernet
> On 3. May 2023, at 19:28, Christian Zigotzky wrote:
>
>
>
>>> On 3. May 2023, at 18:51, Rob Herring wrote:
>>>
>>> On Wed, May 3, 2023 at 11:27 AM Christophe Leroy
>>> wrote:
>>>
>>> +Rob as he's the commit's Author.
>>>
Le 03/05/2023 à 17:46, Christian Zigotzky a écrit :
35 matches
Mail list logo
