March 2013 - kernelnewbies@kernelnewbies.org - Kernelnewbies Mailing Lists

Implementing Network Packet Capturing in Linux-3.5.7
by Prasad Lakshman March 13, 2013

March 13, 2013

Hi, I am trying to implement packet capturing in Linux-3.5.7. I have the procedure which works in Linux-2.4 , will the same procedure work on Linux-3.5.7 I am bit new to this , the procedure i am following is implementing Netfilter hooks capture the packets at several points of path in TCP/IP protocol stack NF_IP_PRE_ROUTING 1st hook NF_IP_LOCAL_IN 2nd hook NF_IP_FORWARD 3rd NF_IP_POST_ROUTING 4th NF_IP_LOCAL_OUT 5th Each hook is activated. Kernel modules can register to listen at any of these hooks. 1. NF_ACCEPT . 2. NF_DROP . 3. NF_STOLEN . 4. NF_QUEUE . 5. NF_REPEAT . does this procedure work with the new kernel version also now from user space i am going to use Libipq I for communicating with ip_queue. Please someone let me know if i am missing something, Regards, Blprasad.

1 0

BUG: unable to handle kernel paging request EIP is at kmem_cache_alloc when testing a new bluetooth driver
by Ahmed Abdelfattah March 13, 2013

March 13, 2013

Hello , I am writing a kernel module to abstract a Bluetooth USB dongle as a character device (using btusb as a base for my work) so that the user space bluetooth stacks (like the new broadcom stack released in Android 4.2) can communicate with the Bluetooth USB device. I tested the driver by a writing simple HCI reset command and monitoring the HCI event my driver works OK for a short time but then the kernel randomly panics after undetermined number of loops writing reset and reading an event . The EIP is always at kmem_cache_alloc but with different stack traces. my guess is that I am somehow corrupting the memory but I can't really determine how to debug because the kernel doesn't break in my code but as I mentioned before I get usually a different stack trace. Here is is an example pastebin.com/9jhUuaDn and another log pastebin.com/Ftzju4BZ Any idea what can cause this panic or how to debug it ? -- regards , Ahmed Abdelfattah

1 0

pthread_lock
by ishare March 13, 2013

March 13, 2013

when pthread_unlock(mutex) is called , do other threads waiting for mutex be waked up immediately? or be waked up at the next schedule?

4 7

User space memory
by Christoph Seitz March 13, 2013

March 13, 2013

Hi all, I have some problems allocation Memory the right way and use it in my kernel module. I use a char device for reading and writing from/to a pcie dma card. Especially the read function makes me some headache. The user allocates some memory with posix_memalign and call the read function on the device, so that the devices knows where to write to. My driver now uses get_user_pages() to pin the user pages. The memory has never been written or read by the user, so it's not yet in the RAM, right? And get_user_pages returns a valid number of pages, but for every page the same struct. (respectively the same pointer). Is there any way to ensure that the user pages are in the ram and get_user_pages returns a valid page array? Regards Chris

3 5

Is this normal or wrong?
by ishare March 13, 2013

March 13, 2013

In init procedure ,I use kmem_cache_alloc to apply for memory to locate a inode struct. it returns to me a inode whose i_dentry element cantain a struct list_head pointing to address 0x3, Is the address within low momery resonable or illegal? thanks!

2 2

Difference between task_active_pid_ns(current) and current->nsproxy->pid.
by Raphael S Carvalho March 12, 2013

March 12, 2013

1169 /* 1170 * If the new process will be in a different pid namespace 1171 * don't allow the creation of threads. 1172 */ 1173 if ((clone_flags & (CLONE_VM|CLONE_NEWPID)) && 1174 (task_active_pid_ns(current) != current->nsproxy->pid_ns)) 1175 return ERR_PTR(-EINVAL); Navigating through these calls, I discovered that task_active_pid_ns calls task_pid(), which gets the struct pid attached to the task_struct, and use this struct as an argument to ns_of_pid() ( E.G: return ns_of_pid(task_pid(tsk)) ). Reading the code carefully, I realized that ns_of_pid() returns the higher level pid_namespace attached to the struct pid. Thus, I think that the higher level pid_namespace in the struct pid will always be the active pid_namespace. When would task_active_pid_ns(current) be different than current->nsproxy->pid_ns? Sorry if this question was already asnwered somewhere, I didn't find it. Regards, Raphael S. Carvalho.

1 0

Explanation of <asm/gpio.h> in Doc/gpio.txt seems out of date
by Robert P. J. Day March 12, 2013

March 12, 2013

(the first in probably a number of questions related to digging around in embedded linux-related code -- i know there's an embedded-linux list, if that's a better forum, i'll move stuff there.) currently writing an intro to GPIO for my new EL course, and ran across this in doc file gpio.txt: "Platforms must declare GENERIC_GPIO support in their Kconfig (boolean true), and provide an <asm/gpio.h> file. Drivers that can't work without standard GPIO calls should have Kconfig entries which depend on GENERIC_GPIO. The GPIO calls are available, either as "real code" or as optimized-away stubs, when drivers use the include file: #include <linux/gpio.h>" hang on ... that explanation might have been true at one time, but consider this snippet from <linux/gpio.h>: ... snip ... #ifdef CONFIG_GENERIC_GPIO #ifdef CONFIG_ARCH_HAVE_CUSTOM_GPIO_H #include <asm/gpio.h> #else #include <asm-generic/gpio.h> ... snip ... the above clearly suggests that you need provide an <asm/gpio.h> only if you've explicitly selected that you have a custom gpio.h file, and that Kconfig directive is defined thusly in drivers/gpio/Kconfig: config ARCH_HAVE_CUSTOM_GPIO_H bool help Selecting this config option from the architecture Kconfig allows the architecture to provide a custom asm/gpio.h implementation overriding the default implementations. New uses of this are strongly discouraged. so, if i read this correctly, you *don't* need to supply your own <asm/gpio.h> except under exceptional circumstances which, these days, is strongly discouraged, does that sound right? further, there are numerous kernel source files that explicitly have: #include <asm/gpio.h> when it would *seem* that the cleaner approach would be to simply: #include <linux/gpio.h> and set the Kconfig variable ARCH_HAVE_CUSTOM_GPIO_H, no? it just seems that the Doc/gpio.txt file could stand a bit of tweaking to be brought up to date. rday

1 0

pi_waiters in task_struct
by Kevin Wilson March 12, 2013

March 12, 2013

Hi, Looking into task_struct (include/linux/scehd.h) , we have: /* PI waiters blocked on a rt_mutex held by this task */ struct plist_head pi_waiters; Any ideas what "pi" stand for ? I assume it has nothing to do with "pi" from mathematics... rgs. Kevin

2 1

location of inbuilt system call implementation
by Niroj Pokhrel March 11, 2013

March 11, 2013

Hi All, I am newbies to Linux and am going through system call in kernel-3.4. I just wanted to check how sys_read, sys_write or other system calls work but I was searching for the implementation but couldn't find it. I found that : System Call are added in /arch/x86/syscalls/syscall_32(or64).tbl. and they were declared in /incude/linux/sycalls.h But I was not able to find exactly where they are implemented. Thanking you all in advance.

2 1

the most socket num of epoll
by ishare March 11, 2013

March 11, 2013

In one process , can I set the num of socket that is managed by epoll to 4096 or bigger ? thanks!

1 0