August 2021 - kernelnewbies@kernelnewbies.org

How to implement waiting until processing is done by a device..
by ckim＠etri.re.kr Aug. 20, 2021

Aug. 20, 2021

Hello, We are developing a compute accelerator and I'm not sure which is the best or normal way to implement the waiting method for the application until the computing is finished by the accelerator. The application first sets up input data and configuration for the accelerator on the memory. It then calls an ioctl which sets up the registers and it calls another ioctl which trigger the accelerator. My idea is : * The application process goes into wait queue after calling the second ioctl by telling the OS "keep me in the wait queue and wake me up when 'this' signal comes'. * The driver's ISR process sends a signal to the sleeping user process when processing the 'computing complete' interrupt. Does this make sense? (Can I define and use a signal for this purpose?) And if it does, what function calls should I use? I would appreciate sending me any link for suggested reading. It should be possible to use accelerator for multiple processes at the same time, but I think for now we can just do with single application. Thanks! Chan Kim

1 0

Eudyptula Challenge.
by David K Aug. 20, 2021

Aug. 20, 2021

Hello, It is my plan to share some, if not all, of my material from Eudyptula challenge here mostly with the intent of getting critique. This should not be against any Eudyptula challenge rules as a lot of people have already shared their solutions to the challenge on Github and probably other places. Since email is the accepted way of contributing to Linux kernel development and a lot of other open source projects, this is my preferred way of sharing my solutions and possibly getting critique. It is my hope that this is not against any list rules. This looks like the case to me and would it happen that this is not the case, any heads up would be much appreciated. Regards, David.

2 1

Re: kmalloc - Address is not consecutive
by Lloyd Aug. 19, 2021

Aug. 19, 2021

Thanks a lot, it worked :) On Thu, Aug 19, 2021 at 7:00 PM ckim <ckim(a)etri.re.kr> wrote: > Hi! this one is easy :) > For printing pointer in kernel space, you should use %px not %p as format > specifier. I heard this is for security reason(?). > Chan Kim > > 내 Galaxy에서 보냄 > > > -------- 원본 이메일 -------- > 발신: wuzhouhui <wuzhouhui14(a)mails.ucas.ac.cn> > 날짜: 21/8/19 오후 9:17 (GMT+09:00) > 받은 사람: Lloyd <lloydkl.tech(a)gmail.com>, kernelnewbies(a)kernelnewbies.org > 제목: Re: kmalloc - Address is not consecutive > > > > On 8/19/21 6:54 PM, Lloyd wrote: > > Hi, > > > > I am trying to print the address of consecutive locations in a > > memory allocated using kmalloc. It prints non contiguous addresses! I > > tried with kmalloc_array, and normal array on stack, I got the same > > result. May I know the reason? The code fragment for array and result > > are given below. > > > > uint8_t MyArray[100]={'0'}; > > printk(KERN_INFO "\nAddress is-> %p\n",&MyArray[0]); > > printk(KERN_INFO "\nAddress is-> %p\n",&MyArray[1]); > > printk(KERN_INFO "\nAddress is-> %p\n",&MyArray[2]); > > > > > > [ 8684.613909] Address is-> 00000000b3330991 > > [ 8684.613914] Address is-> 000000000519068c > > [ 8684.613915] Address is-> 00000000e448ca76 > > > Quoted from Documentation/core-api/printk-formats.rst: > > A raw pointer value may be printed with %p which will hash the address > > before printing. The kernel also supports extended specifiers for > printing > > pointers of different types. > So the address printed neither logical nor physical address. > > > Thanks a lot, > > Lloyd > > > > > > _______________________________________________ > > Kernelnewbies mailing list > > Kernelnewbies(a)kernelnewbies.org > > https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies > > > _______________________________________________ > Kernelnewbies mailing list > Kernelnewbies(a)kernelnewbies.org > https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies > _______________________________________________ > Kernelnewbies mailing list > Kernelnewbies(a)kernelnewbies.org > https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies >

1 0

kmalloc - Address is not consecutive
by Lloyd Aug. 19, 2021

Aug. 19, 2021

Hi, I am trying to print the address of consecutive locations in a memory allocated using kmalloc. It prints non contiguous addresses! I tried with kmalloc_array, and normal array on stack, I got the same result. May I know the reason? The code fragment for array and result are given below. uint8_t MyArray[100]={'0'}; printk(KERN_INFO "\nAddress is-> %p\n",&MyArray[0]); printk(KERN_INFO "\nAddress is-> %p\n",&MyArray[1]); printk(KERN_INFO "\nAddress is-> %p\n",&MyArray[2]); [ 8684.613909] Address is-> 00000000b3330991 [ 8684.613914] Address is-> 000000000519068c [ 8684.613915] Address is-> 00000000e448ca76 Thanks a lot, Lloyd

3 2

[RFC PATCH v3 00/18] Add block device LED trigger
by Ian Pilcher Aug. 19, 2021

Aug. 19, 2021

Ridiculous or not, here is version 3 of the block device trigger for "freaking blinkenlights". It addresses basically all of the points raised in response to the v2 patchset. * The main body of the code is moved from the block directory into the LED triggers directory (drivers/leds/trigger/ledtrig-blkdev.c) The downside of this is that it requires adding an API to the block subsystem - get_disk_by_name() - which allows the trigger code to resolve a gendisk when asked to monitor it. I know of no good way to do this today, and I don't know of a good way to implement the sysfs API requested by Pavel and Marek without something like this API. Other than that, changes to the block subsystem are as minimal as I can make them - a single pointer added to struct gendisk and init/cleanup calls when a gendisk is added or deleted. * This also implements Marek's suggestion of periodically checking devices for activity, rather than directly blinking LEDs in the I/O path. This change has the unanticipated benefit of making the trigger work on pretty much all types of virtual block devices (device mapper, MD RAID, zRAM, etc.), as well as NVMe SSDs. * Relationships between devices and LEDs are now many-to-many. An LED can monitor multiple devices, and multiple LEDs can monitor any one device. The current "associations" are reflected in two sysfs directories. - /sys/class/leds/<led>/block_devices contains links to all devices associated with an LED, and - /sys/block/<disk>/blkdev_leds contains links to all LEDs with which the device is associated. (The latter directory only exists when the device is associated with at least one LED.) * Each LED can be set to show read activity, write activity, or both. Discards and cache flushes are considered to be writes, as they affect the state of the device's non-volatile storage. Ian Pilcher (18): docs: Add block device (blkdev) LED trigger documentation block: Add get_disk_by_name() for use by blkdev LED trigger ledtrig-blkdev: Add file (ledtrig-blkdev.c) for block device LED trigger ledtrig-blkdev: Add misc. helper functions to blkdev LED trigger ledtrig-blkdev: Periodically check devices for activity & blink LEDs block: Add LED trigger pointer to struct gendisk ledtrig-blkdev: Add function to initialize gendisk ledtrig member ledtrig-blkdev: Add function to remove LED/device association ledtrig-blkdev: Add function to disassociate a device from all LEDs block: Call LED trigger init/cleanup functions ledtrig-blkdev: Add function to associate a device with an LED ledtrig-blkdev: Add sysfs attributes to [dis]associate LEDs & devices ledtrig-blkdev: Add blink_time & interval sysfs attributes ledtrig-blkdev: Add mode (read/write/rw) sysfs attributue ledtrig-blkdev: Add function to associate blkdev trigger with LED ledtrig-blkdev: Add function to disassociate an LED from the trigger ledtrig-blkdev: Add initialization function ledtrig-blkdev: Add config option to enable the trigger Documentation/ABI/testing/sysfs-block | 9 + .../testing/sysfs-class-led-trigger-blkdev | 48 ++ Documentation/leds/index.rst | 1 + Documentation/leds/ledtrig-blkdev.rst | 132 +++ block/genhd.c | 28 + drivers/leds/trigger/Kconfig | 9 + drivers/leds/trigger/Makefile | 1 + drivers/leds/trigger/ledtrig-blkdev.c | 770 ++++++++++++++++++ include/linux/genhd.h | 13 + include/linux/leds.h | 20 + 10 files changed, 1031 insertions(+) create mode 100644 Documentation/ABI/testing/sysfs-class-led-trigger-blkdev create mode 100644 Documentation/leds/ledtrig-blkdev.rst create mode 100644 drivers/leds/trigger/ledtrig-blkdev.c -- 2.31.1

1 18

kmalloc - Crashing
by Lloyd Aug. 18, 2021

Aug. 18, 2021

Hi, I am starting to write linux kernel mode code. I am on Ubuntu 20.04. I have written the following code The line DynPage[i]=0; //CRASH causes a crash. Can you please guide on the reason for the crash? Thanks a lot in advance, Lloyd #pragma GCC push_options #pragma GCC optimize("O0") void ClearMem(void) { unsigned int* DynPage=NULL; unsigned int i=0; DynPage=kmalloc(1024*16,GFP_KERNEL); if(DynPage!=NULL) { DynPage[i]=0; //CRASH } kfree(DynPage); DynPage=NULL; return; } #pragma GCC pop_options

2 3

Process after patch has been reviewed and reviewed-by: tag added
by nishad kamdar Aug. 18, 2021

Aug. 18, 2021

Hello, I submitted a patch here: https://patchwork.kernel.org/project/linux-mmc/patch/20210812065730.3986-1-… It has been reviewed and the reviewed-by: tag has been added. But since then it has been a week and there is no activity. I wanted to know what happens after this step and is the patch ready for acceptance ? Should I remind some of the maintainers to look into this patch or is there nothing to be done from my end? Thanks for your time and any suggestions would be appreciated. Thanks and Regards, Nishad

2 2

"Forcibly" disabling a config option (CONFIG_BLOCK)?
by Ian Pilcher Aug. 17, 2021

Aug. 17, 2021

I need to test a kernel build with the block layer (CONFIG_BLOCK) disabled. Starting with a realistic configuration, I haven't found a way to disable it, as it's selected by so many other options. Is there any convenient way to "forcibly" disable an option (along with everything that depends on or selects it)? TIA! -- ======================================================================== In Soviet Russia, Google searches you! ========================================================================

1 0

Problem for testing linux-next
by Sidong Yang Aug. 16, 2021

Aug. 16, 2021

Hi, all. I'm trying to test linux-next. I'm using QEMU for testing. I cloned linux-next repository and checkout to last tag next-20210812. I compiled with simple command below. make menuconfig make & make modules_install & make install and rebooted system. and it failed. It hangs after the message "freeing unused kernel image (init mem) memory" after about 500 seconds, it prints call trace including "x2apic_send_IPI" and it blocked. I want to copy and past it to here. but I don't know how to see the dmesg in last failed session. Is there any wrong configuration for compiling? I would paste it if I could copy the error message. Thanks, Sidong

2 2

kernel / system configuration question regarding "unstable filesystems"
by Colin Williams Aug. 16, 2021

Aug. 16, 2021

Hello, I'm using a SD card to boot my device. The SD card is connected via a USB-C hub due to the device not having any other connectors. This hub and slot will lose power temporarily if plugging in or unplugging power passthrough to the hub. Then on these events the filesystem becomes unmounted and while my system is still running, it becomes unusable. I get input / output errors for example from the terminal. I think what's going on is some portion of the system is loaded into memory. However many critical parts become unavailable. Then I'm curious what options I have at the kernel or system level to ruggedize this so these events don't destroy the system. Before exposing my own ignorance presenting any ideas with my limited knowledge regarding this topic. Is anyone familiar enough to suggest some design approaches or where I might find the best answer regarding a designed solution?

1 0