December 2017 - kernelnewbies@kernelnewbies.org

Shared DDR RAM access on AMP system
by Edward Wingate Dec. 14, 2017

Dec. 14, 2017

On a dual-core ARM SOC (Zynq7020), I have Linux running on core0 and an RTOS running on core1. I share data between the two cores with a 4MB block of DDR RAM that I've marked in the MMU as shareable and non-cacheable. I use an entry in the device tree and the uio_pdrv_genirq driver to create an uio device (/dev/uio0) in Linux. With this device, I can use mmap to read/write the block of DDR RAM from Linux. Even though it's meant for memory mapped devices, the uio_pdrv_genirq driver seems to work for accessing shared DDR RAM. Are there any potential issues that may arise from using this driver for accessing shared DDR RAM? Is the uio_dmem_genirq driver more appropriate for this? I couldn't get it to work though, perhaps because of this bug: https://forums.xilinx.com/t5/Embedded-Linux/Bug-in-uio-dmem-genirq-UIO-driv… Thanks for any advice/help. Ed

1 0

hrtimer_interrupt time sync issues across cores
by Rajasekaran Chandrasekaran Dec. 14, 2017

Dec. 14, 2017

Hi, In our multi-core x86 based system that is running 3.4.19 version of kernel, hrtimer_interrupt (called from apic_timer_interrupt) keeps looping in hardirq for atleast 1.6 seconds. We use tsc as our clock source. The issue happens very rarely in our system and hard to reproduce. Problem: Inside hrtimer_interrupt function, basenow.tv64 in CPU-3 is 1.6 seconds ahead of other CPU’s (we have 4 cores), whereas hrtimer->_softexpires.tv64 is in sync with remaining CPU’s. Due to this, the if condition inside hrtimer_interrupt where we check if basenow.tv64 < hrtimer_get_softexpires_tv64(timer) is not true for 1.6 seconds, which cause the while loop inside hrtimer_interrupt to not exit. Below is the ftrace captured during the problem. <idle>-0 [002] d.h. 800364.533632: hrtimer_expire_entry: hrtimer=ffff88017fd0c960 function=tick_sched_timer now=801616439840902 ksoftirqd/3-19 [003] dNh. 800364.539178: hrtimer_expire_entry: hrtimer=ffff88017fd8c960 function=tick_sched_timer now=801618042768641 ksoftirqd/3-19 [003] dNh. 800364.539185: hrtimer_start: hrtimer=ffff88017fd8c960 function=tick_sched_timer expires=801616446505014 softexpires=801616446505014 As we can see, the difference in now time between CPU-2 and CPU-3(where the time jump is seen) is significant. Ftrace indicates that the now time has drifted apart in CPU-3 by 1602 milliseconds, even though timestamp is apart by only 6 milliseconds. Also since the hrtimer expiry time is in the past, we end up spending lot of time in hardirq. From my understanding of the code, , basenow.tv64 is computed in hrtimer_update_base() ->ktime_get_update_offsets() as timekeeper.xtime – offs_real. Both timekeeper.xtime and offs_real are always updated under a lock. So, I am still unsure on how only one core is seeing the time incorrectly. Any inputs will be greatly help. Thanks, Raj

3 2

Virtual Memory map
by Damian Tometzki Dec. 13, 2017

Dec. 13, 2017

Hello all, i have a short question: How are the bits (in brackets) calculated in the virtual memory map ? for example the first line (=47 bits) Virtual memory map with 4 level page tables: 0000000000000000 - 00007fffffffffff (=47 bits) user space, different per mm hole caused by [47:63] sign extension ffff800000000000 - ffff87ffffffffff (=43 bits) guard hole, reserved for hypervisor Best regards Damian

3 4

Virtual Memory Map
by Damian Tometzki Dec. 12, 2017

Dec. 12, 2017

Hello all, i have a short question: How are the bits (in brackets) calculated in the virtual memory map ? for example the first line (=47 bits) Virtual memory map with 4 level page tables: 0000000000000000 - 00007fffffffffff (=47 bits) user space, different per mm hole caused by [47:63] sign extension ffff800000000000 - ffff87ffffffffff (=43 bits) guard hole, reserved for hypervisor Best regards Damian

1 0

Re: sched_child_runs_first doesn't work
by strongbox8 Dec. 11, 2017

Dec. 11, 2017

2 1

Header organization/separation rule in kernel
by Shiyao MA Dec. 10, 2017

Dec. 10, 2017

Hi, I wonder the header separation rule in kernel. To make this question concrete, for example, let's targeting genetlink.h It can be found in: http://elixir.free-electrons.com/linux/latest/source/include/net/genetlink.h http://elixir.free-electrons.com/linux/latest/source/include/linux/genetlin… http://elixir.free-electrons.com/linux/latest/source/include/uapi/linux/gen… so basically three parts, in uapi, linux/genetlink.h, and net/genetlink.h. I understand the necessity of uapi, but why divide the genetlink.h in linux and the net folder? What's the benefit of such separation? -- Best, Shiyao

3 2

Re: sched_child_runs_first doesn't work
by Perr Zhang Dec. 9, 2017

Dec. 9, 2017

maybe you should run the program like this: taskset 1 ./your_test_app

3 2

Re: sched_child_runs_first doesn't work
by Perr Zhang Dec. 8, 2017

Dec. 8, 2017

maybe you should run the program like this:

1 0

how to hard hang the kernel from userspace?
by Daniel Walker Dec. 8, 2017

Dec. 8, 2017

Hi, I was able to formulate this command, echo -1 > /proc/sys/kernel/sched_rt_runtime_us chrt -f 99 yes &> /dev/null and it works sometimes on UP, but it doesn't work on SMP. I also tried adding in "taskset" to pin multiple version of the "yes" command to each cpu in the system, but that just resulted in lots of rcu warnings. The goal is to make the watchdog trigger by starving out the watchdog petting process. Anyone have any ideas? I'd rather not make a kernel module. Thanks, Daniel

2 4

How to tell if a field in c struct is initialized?
by Shiyao MA Dec. 8, 2017

Dec. 8, 2017

Hi, I come across this code: http://elixir.free-electrons.com/linux/latest/source/drivers/scsi/scsi_netl… other fields such as .cbmutex, .bind, etc are not initialized. Since the object =cfg= is of automatic storage, I wonder how the kernel tells if a field is initialized or its value is just some random bits on the stack? -- Best, Shiyao

2 2