December 2011 - kernelnewbies@kernelnewbies.org

Issue with ioctl commands - Linux version 2.6.35.7+
by Chetan Nanda Dec. 28, 2011

Dec. 28, 2011

Hi All, I am facing a strange issue with the ioctl commands, and not able to find what wrong I am doing. I am trying with a dummy kernel driver and implemented ioctl command as follow: in my_ioctl.h #define READHWREG_MAGIC_NUMBER 0x15 #define READHWREG_CAM_READPE _IOR(READHWREG_MAGIC_NUMBER, 1, int*) But the value of 'READHWREG_CAM_READPE' is comes out to be different in userspace (in the application) and kernel space (in the driver). In userspace I used ioctl as ... printf("\ncommand %x \n",READHWREG_CAM_READPE); <-- command 80041501 error = ioctl(g_DevFileId, READHWREG_CAM_READPE); .. In kernel space int readHwReg_ioctl(struct inode *node, struct file *filp, unsigned int cmd, unsigned long arg) { ... printk("\n%d %s cmd = %x\n",__LINE__,__FUNCTION__,cmd); <--- cmd = bee6a9d4 ... } I am not able to figure out the root cause, Please help to check... Thanks, Chetan Nanda

3 3

Message signal interrupts: Rule understanding
by Mukesh Yadav Dec. 27, 2011

Dec. 27, 2011

Hi Guys, Want to know some details about MSI queues. Looked into many only online references for MSI, they points mainly to introduction area. I am looking for rule/decision criteria for distributing traffic on different MSI queues. Can we configure MSI queue at runtime for traffic distribution. On my machine I can see below data in /proc/interrupts 64: 218601 13334 25585 7505 PCI-MSI-edge eth0-0 65: 5717754 6491556 501052 729091 PCI-MSI-edge eth0-1 66: 844740 9897919 96 230 PCI-MSI-edge eth0-2 67: 222750 436800 403 1205846 PCI-MSI-edge eth0-3 68: 777 1502281 314536 100125 PCI-MSI-edge eth0-4 69: 482616 431247 2164970 1627540 PCI-MSI-edge eth0-5 70: 323501 1433873 81970 18359 PCI-MSI-edge eth0-6 71: 37298 35844 8271 18516 PCI-MSI-edge eth0-7 When send UDP packets with IP X.X.X.100 interrupt happens in eth0-1 and with IP X.X.X.101 interrupts happens in eth0-5. Similary for Ip X.X.X.102 and X.X.X.104, interrupt happens on distinct MSI queue...Seems like modules 4 operation here being criteria. Can we have vlan based MSI queue rules? When I set the core affinity, with /proc/irq/<intrrpt no>/smp_affinity. It changes on its own when I send burst of UPD packets with above IP as dest. Is this expected? Thanks Mukesh

1 0

Re: Issue with ioctl commands - Linux version 2.6.35.7+
by Chetan Nanda Dec. 27, 2011

Dec. 27, 2011

On 12/27/11, Alexandru Juncu <alex.juncu(a)rosedu.org> wrote: > On Tue, Dec 27, 2011 at 1:57 PM, Chetan Nanda <chetannanda(a)gmail.com> wrote: >> >> >> On Tue, Dec 27, 2011 at 5:23 PM, Alexandru Juncu <alex.juncu(a)rosedu.org> >> wrote: >>> >>> On Tue, Dec 27, 2011 at 1:39 PM, Chetan Nanda <chetannanda(a)gmail.com> >>> wrote: >>> > Hi All, >>> > >>> > I am facing a strange issue with the ioctl commands, and not able to >>> > find >>> > what wrong I am doing. >>> > I am trying with a dummy kernel driver and implemented ioctl command as >>> > follow: >>> > >>> > in my_ioctl.h >>> > #define READHWREG_MAGIC_NUMBER 0x15 >>> > #define READHWREG_CAM_READPE _IOR(READHWREG_MAGIC_NUMBER, 1, int*) >>> > >>> > But the value of 'READHWREG_CAM_READPE' is comes out to be different in >>> > userspace (in the application) and kernel space (in the driver). >>> > >>> > In userspace I used ioctl as >>> > ... >>> > printf("\ncommand %x \n",READHWREG_CAM_READPE); <-- command 80041501 >>> > error = ioctl(g_DevFileId, READHWREG_CAM_READPE); >>> > .. >>> > >>> > In kernel space >>> > >>> > int readHwReg_ioctl(struct inode *node, struct file *filp, unsigned int >>> > cmd, unsigned long arg) >>> > { >>> > ... >>> > printk("\n%d %s cmd = %x\n",__LINE__,__FUNCTION__,cmd); <--- cmd = >>> > bee6a9d4 >>> > ... >>> > } >>> > >>> > >>> > I am not able to figure out the root cause, >>> > Please help to check... >>> >>> I think that the command is actually sent via the arg parameter, which >>> is an address (it's a pointer to the value your are sending to kernel >>> space). >> >> >> Yes, even I observe this now. But how this can happen. >> 'command' is the second argument for ioctl call. > > As far as I remember, you can send an integer through the cmd, but > other things (like structures or buffers) through their address (and > copy_from_user-ed). But, maybe, on new implementations (or on some > architectures) everything is send via address. What I have observed, third argument of the ioctl system call is reaching the driver as a command. ioctl(g_DevFileId, READHWREG_CAM_READPE, xx); In driver: int readHwReg_ioctl(struct inode *node, struct file *filp, unsigned int cmd, unsigned long arg) cmd is equal to xx Am I missing something in the driver/application. Thanks, Chetan Nanda

1 0

Strange kernel reboot (oops?)
by Ezequiel García Dec. 26, 2011

Dec. 26, 2011

Hi, While running an embedded linux platform (sh4) with an old linux kernel (2.6.23), I'm trying to debug a very strange behavior (at least strange to me). The box is running aparently fine, and out of nothing it is rebooting. There is no "Caught signal ..." message, no kernel oops message, just a plain cpu reboot. After the reboot, the linux kernel boots but somewhere before running the user application it panics. The kernel panic message seems useless, because the system is booting. When power is unplugged and re-plugged, the system starts normally. Has anyone seeing this kind of behavior before? Thanks, Ezequiel.

2 1

A query about typeof
by contemplating zombie Dec. 24, 2011

Dec. 24, 2011

Hi, I see following code in kfifo.h typeof((fifo) + 1) __tmp = (fifo); [http://lxr.linux.no/linux+v3.1.6/include/linux/kfifo.h#L334] I do not understand this specific code construct. Why is it necessary to add 1 to fifo before getting its type? Is it related to some 'C' magic? -ctmzb

2 1

Packet drop while using BPF filter
by Mukesh Yadav Dec. 24, 2011

Dec. 24, 2011

Hi Guys, I am working on user level code which uses BPF filter. Solution has two threads and BPF filter for incoming packets to raw socket dedicated to each thread. Intent is to divide ingress traffic between two thread based on whether dest IP is even or odd. BPF filters used are: Even IP filter :-> tcpdump -i interface 'ether dst <dest mac> && ip[19] & 0x01 = 0' Odd IP filter :-> tcpdump -i interface 'ether dst <dest mac> && ip[19] & 0x01 = 1' There is a packet loss for odd IP thread even when CPU is available. Reason being packet drop due recv buffer full. Same amount of traffic is well handled by even IP thread(user code being same in both), scaling CPU to full 100%. In odd IP thread, If filter is changed to "tcpdump -i interface 'ether dst <dest mac> && ip proto 17' ", all goes fine. Also CPU usage at kernel drops from 50% to 4 % for a particular amount of traffic. Would appreciate any I/ps for reducing load at kernel for packet filtering(Odd/Even IP distribution) Cheers Mukesh

3 6

TCP checksum offloading for raw sockets
by Mukesh Yadav Dec. 24, 2011

Dec. 24, 2011

Hi Guys, How to use TCP checksum offloading for raw sockets. I can see rx/tx checksumming as enabled using ethool. But packets send out from machine have IP/UDP checksum as zero only. While making packet I am making checksum value in IP/UDP header as zero. No new value updated in checksum. >ethtool -k eth0 Offload parameters for eth0: rx-checksumming: on tx-checksumming: on scatter-gather: on tcp-segmentation-offload: on udp-fragmentation-offload: off generic-segmentation-offload: on generic-receive-offload: off large-receive-offload: off -- Cheers Mukesh

1 0

delay in booting due to consloe_unlock inprintk
by Syam Dec. 24, 2011

Dec. 24, 2011

HI All, I am getting more than 1 sec delay during booting. While registering console (for serial console ttys ), there is about 1 sec delay in executing the console_unlock() in printk.c It was found that delay is occuring in call_console_drivers() ****************************** if (c == '\n') { if (msg_level < 0) { /* * printk() has already given us loglevel tags in * the buffer. This code is here in case the * log buffer has wrapped right round and scribbled * on those tags */ msg_level = default_message_loglevel; } _call_console_drivers(start_print, cur_index, msg_level); // if this part is commented, there is no delay observed? ******************************* is there any way to find out what is causing the delay. -- Warm Regards, Syam

3 3

Metadata snapshot issues
by Swapnil Gaikwad Dec. 23, 2011

Dec. 23, 2011

If we gives new inodes to each file during metadata snapshot. Then is there any conflicting issue happens? What are the techniques help us in this? Is some-one have source code of it? -- Regards, Swapnil Gaikwad.

2 1

kernel lock-free stack
by srimugunthan dhandapani Dec. 23, 2011

Dec. 23, 2011

Hi all, Are there any example implementations in the linux kernel for lock free stacks,queues and linked list using "Compare and swap" primitive? Thanks, mugunthan

1 0