September 2016 - kernelnewbies@kernelnewbies.org

scatter-gather doubt?
by Muni Sekhar Sept. 26, 2016

Sept. 26, 2016

Hi All, I am working on a xilinx PCIe endpoint with DMA reference block. The DMA reference block design has 2 Scatter-Gather engines, one for each DMA channel. Channel 0 is for HostMemory -> DMA_REF FIFO transfers Channel 1 is for DMA_REF FIFO -> HostMemory transfers Each scatter-gather engine works through a linked list of Descriptors from which it generates the required DMA activity. The format of these descriptors is depicted as below: Offset @ 0x00 - LSBs of pointer to DMA data Offset @ 0x04 - MSBs of pointer to DMA data Offset @ 0x08 - Number of data bytes to be transferred. (note: only 8 byte aligned transfers supported) Offset @ 0x0C - LSBs of pointer to next Descriptor (Set this field & MSBs to zero to indicate end of descriptor list) Offset @ 0x10 - MSBs of pointer to next Descriptor Does the Linux kernel has any data structure to support the above mentioned scatter-gather descriptor? Will it be possible to use the kernel scatterlist API’s for this hardware? -- Thanks, Sekhar

2 2

FTrace
by Dave Frost Sept. 24, 2016

Sept. 24, 2016

Hi Folks, Im trying to get a handle on how the Trace meta data collection works - i can roughly get a grasp of what the SYSCALL_METADA macro does but there are a few curious compiler flags that I've never used before, and think i need some help understanding their implications. Ive been looking in vim include/linux/syscalls.h and look at the macros SYSCALL_TRACE_ENTER_EVENT SYSCALL_TRACE_EXIT_EVENT SYSCALL_METADATA and these macros create some data structures, but don't seem to call any functions to make any use of them, so I'm assuming that these macros set up some data structures used by something else. They all end with an assignment that has a compiler directive like __attribute__((section("xxxxxx"))) for example, SYSCALL_METADATA ends with static struct syscall_metadata __used \ __attribute__((section("__syscalls_metadata"))) \ *__p_syscall_meta_##sname = &__syscall_meta_##sname; This data structure has been been placed into a special section in the kernel called "__syscalls_metadata" Is this a heap kind of area somewhere ? i assume that something is watching this area for changes (or being signal somehow) so it will see these changes in data ? the macros SYSCALL_TRACE_ENTER_EVENT have similar attributes on their assignment that put the data into "_ftrace_events". I suppose trying to follow logically through the code has now put me into a bit of a dead end - can anyone point me to the documentation / code that reads this data so i can complete the chain execution ? Cheers Dave

1 0

How to search topics in The Kernelnewbies Archives?
by Chi Wang Sept. 24, 2016

Sept. 24, 2016

Hello everyone, I have just subscribed to this mail list, and trying browsing history topics in this page: https://lists.kernelnewbies.org/pipermail/kernelnewbies/ Is there a general way to do searching in this archive? Or I have to download all the packages? Regards, Chi Wang

5 4

SMP on different cores.
by Madhu K Sept. 24, 2016

Sept. 24, 2016

Hi All, If i want to enable SMP, cores should be of same version( ex: both core should ARM 7 ) or we could enable SMP on different version. Thanks Madhu

5 6

What causes a recv syscall for a socket to sometimes return less than the request len?
by Daniel Ellis Sept. 23, 2016

Sept. 23, 2016

Howdy, I understand that data fragmentation happens across TCP packets, and in my case that is what is happening. The packets I'm seeing usually run around 1450 in length, and are always cut into two packets at the 1400 mark. This works fine most of the time, since the call to recv is being passed a len of 4096 (via Python's socket.recv, which I'm assuming is just a wrapper around the system call). But sometimes I'll get all of the data I request, even across TCP packet boundaries (so the full 1450 bytes), and other times, it stops at a single packet, at which point I need to make another recv call. This seems to be happening more when the box is under more load (both in a network and CPU sense), but that might be a coincidence. Are there some internal buffers that I could be monitoring that would correlate with these packets being flushed out quicker? I've changed the underlying code to correctly call recv until all the data has been received, but I am curious as to what's going on under the hood that made this so sporadic (and a bit of a headache to track down). Any help would be appreciated! Dan

3 2

Re: SMP on different cores.
by Daniel. Sept. 22, 2016

Sept. 22, 2016

s/iMX6/iMX7/ 2016-09-22 10:04 GMT-03:00 Daniel. <danielhilst(a)gmail.com>: > I know about i.MX6 which has Cortex-A7 + Cortex-M4... but I don't > think that M4 is handled by Linux but by some RTOS > > 2016-09-22 9:17 GMT-03:00 Greg KH <greg(a)kroah.com>: >> On Thu, Sep 22, 2016 at 04:46:12PM +0530, Madhu K wrote: >>> Hi All, >>> >>> If i want to enable SMP, cores should be of same version( ex: both >>> core should ARM 7 ) or we could enable SMP on different version. >> >> Do you have a machine that has cores of different versions of a CPU? If >> so, what are they? >> >> thanks, >> >> greg k-h >> >> _______________________________________________ >> Kernelnewbies mailing list >> Kernelnewbies(a)kernelnewbies.org >> https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies > > > > -- > "Do or do not. There is no try" > Yoda Master -- "Do or do not. There is no try" Yoda Master

1 0

RF class driver
by Raul Piper Sept. 21, 2016

Sept. 21, 2016

Hi, I was looking at the different class of the drivers and wanted to know in which class will the RF transciever will fall. Network driver or the IIO driver.How to get the maintainer list name for the same, I ran get_maintainer.pl but couldnt figure out who will be the maintainer for the same. Rgs, Rp

4 9

Changing RT task to a different cgroup
by Rahul Bhattacharya Sept. 20, 2016

Sept. 20, 2016

When I try to move a RT task to a different cgroup(shown below) I get a EINVAL error. However I can change the cgroup first and then change the sched policy. After looking into kernel source code(*I dont use CONFIG_RT_GROUP_SCHED config*) I was able to extract this piece of logic *chrt -p 777* pid 777's current scheduling policy: SCHED_OTHER pid 777's current scheduling priority: 0 *cat /proc/777/cgroup* 2:cpu,cpuacct:/system.slice 1:name=systemd:/system.slice/d.service *chrt -f -p 50 777* root@mgu-high:~# chrt -p 777 pid 777's current scheduling policy: SCHED_FIFO pid 777's current scheduling priority: 50 */bin/echo 777>/sys/fs/cgroup/cpu,cpuacct/interaction.slice/tasks**echo: write error: Invalid argument* After browsing kernel sources I was able to find out the reason of failure. static int cpu_cgroup_can_attach(struct cgroup_subsys_state *css, struct cgroup_taskset *tset) { struct task_struct *task; cgroup_taskset_for_each(task, tset) { #ifdef CONFIG_RT_GROUP_SCHED if (!sched_rt_can_attach(css_tg(css), task)) return -EINVAL; #else */* We don't support RT-tasks being in separate groups */ if (task->sched_class != &fair_sched_class) { return -EINVAL; }* #endif } return 0; } My question is why is this check required? Also, changing cgroups first and then the policy of FIFO works without issue(?) for the task. Thanks br rb

2 2

ndo_start_xmit what to return when packet is too big!?
by Daniel. Sept. 20, 2016

Sept. 20, 2016

Hi everybody, I was looking at network chapter of LDD3. The .ndo_start_xmit member of struct net_device_ops should return netdev_tx_t which are: enum netdev_tx { __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ NETDEV_TX_OK = 0x00, /* driver took care of packet */ NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ }; I'm deling with some awkward hardware where the MTU is only 32 bytes and was wondering. What should that function return in case of "too large packet"?!?!? Best regards, -- "Do or do not. There is no try" Yoda Master

1 0

Difference between config, menuconfig and defconfig
by Madhu K Sept. 19, 2016

Sept. 19, 2016

Hi All, I am Kernel newbie, I want to know the difference between config, menuconfig and defconfig. Thanks, Madhu

4 7