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

question about vlan driver
by Kevin Wilson May 6, 2013

May 6, 2013

Hello, I have a short question about vlan driver; I see in the vlan driver the following code when creating a new vlan interface: vlan_newlink() { ... if (data[IFLA_VLAN_PROTOCOL]) proto = nla_get_be16(data[IFLA_VLAN_PROTOCOL]); else proto = htons(ETH_P_8021Q); ... } see: http://lxr.free-electrons.com/source/net/8021q/vlan_netlink.c#L105 I understand that in the common case, proto is ETH_P_8021Q. How can it be set to a different value when creating a vlan interface ? And to which other values can it be set ? I did not find anything by "vconfig --help", and also by probing into the userspace code of the vconfig tool. any ideas? Regards, Kevin

2 2

when to remount sysfs ?
by ishare May 6, 2013

May 6, 2013

Hi: As i know , when kernel is booting , sysfs is mounted at /sys before the real root fs is mounted . Then , after the real root fs is mounted , when does i need to remount sysfs at the new /sys location ? I know the devfs is remounted by do_mount ("none", "/dev", "devfs", 0, "") , so how is that for sysfs? Thanks!

2 1

Kernel panic when using initramfs instead of initrd
by Ivan Nikolaev May 6, 2013

May 6, 2013

Hello everyone! Briefly, I'm creating a dummy distribution which boots from CD (using VirtualBox for testing purposes) and lives in RAM. It's a very minimalistic kernel which has almost nothing enabled except from core facilities, virtual terminal, framebuffer and serial port. I'm trying to use the "ultimate boot feature" - the initial RAM filesystem. Basically what I have is a plain simple directory with the following contents: ./proc ./sys ./dev ./etc ./etc/inittab ./etc/profile ./etc/passwd ./bin ./bin/init ./bin/sh ./bin/busybox Obviously, at boot /bin/init is executed which is instructed by /etc/inittab to simply run a shell and stay happy. EVERYTHING WORKS if I enable in the kernel some filesystem (like ext2 or squashfs) and create the initrd image of that type and put inside the contents of the directory. Isolinux then boots the kernel this way: LABEL linux SAY Booting linux... KERNEL /vmlinuz APPEND root=/dev/ram0 initrd=/initrd.gz console=ttyS0,38400 vga=0x305 To make this work, I obviously enabled the "Initial RAM filesystem and RAM disk (initramfs/initrd) support" in "General setup" and the "RAM block device support" in "Device drivers - block devices" options. Well, ensuring everything works, I tried to switch to initramfs. Disabled the "RAM block device support", the ext2 and squashfs filesystems and created the init cpio image with a command like this: find | cpio -H newc -o | gzip -9 > initrd.gz The boot arguments are the same. When I boot, what I have is a kernel panic with the following details: List of all partitions: No filesystem could mount root, tried: Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0) Here's my config file: http://pastebin.com/XVd8ZukU Using linux v3.9.0 -- Ivan Nikolaev

2 1

Circular DMA buffers and userspace vma(s)
by Ilyes Gouta May 4, 2013

May 4, 2013

Hi, (userspace DMA) I'm writing a device driver for a h/w decoder which (dma) source buffers should (preferably) be visible to userspace (mmap'ed) so that the application can directly load data into these. The h/w expects the buffer to be circular in physical ram (ring buffer) with data starting from the middle of it, whereas userspace would be providing a vma when mmap()'ing which is linear in its virtual address space. Is it possible to alter the incoming vma so that a first region of that vma is mapped to the bottom of the dma buffer whereas a second region of the vma is mapped to the beginning of the dma buffer? This way the circular layout constraint of the dma buffer won't be visible to the userspace application, while also enabling it to directly load data in the correct order. Is it possible to achieve this (such as via split_vma() and a couple of remap_pfn_range() calls into the mmap handler)? Thanks, Ilyes

1 0

generating the in-kernel "kerneldoc" guides
by Robert P. J. Day May 3, 2013

May 3, 2013

in my experience, a lot of newbies aren't aware of the in-kernel "kerneldoc" documentation so, unsurprisingly, i have a wiki page for it: http://www.crashcourse.ca/wiki/index.php/Kerneldoc enjoy. rday -- ======================================================================== Robert P. J. Day Ottawa, Ontario, CANADA http://crashcourse.ca Twitter: http://twitter.com/rpjday LinkedIn: http://ca.linkedin.com/in/rpjday ========================================================================

1 0

Why kernel reloads these segment registers, such as ds, es, .etc in startup_32 ?
by Jacky May 2, 2013

May 2, 2013

Hi, All, Why kernel reloads these segment registers in startup_32 again ? Just liking the following: ======================= ENTRY(startup_32) movl pa(stack_start),%ecx /* test KEEP_SEGMENTS flag to see if the bootloader is asking us to not reload segments */ testb $(1<<6), BP_loadflags(%esi) jnz 2f /* * Set segments to known values. */ lgdt pa(boot_gdt_descr) movl $(__BOOT_DS),%eax movl %eax,%ds ... ====================== Considering bootloader, such as GRUB has already load these segments, so why kernel reloads them again while wasting CPU ? Thanks, Jacky

1 1

kernel panic in sample block device driver
by Kumar amit mehta May 2, 2013

May 2, 2013

Hi, I'm new to block layer in linux and to learn the same, I'm trying to come up with a sample memory based block device driver, with which I can experiment and learn along the way. I'm referring to sample code from the linux tree [1] and assorted information available over the internet. My current module is causing system crash as soon I load it. Please take a look. #include <linux/kernel.h> #include <linux/module.h> #include <linux/sched.h> #include <linux/blkdev.h> #include <linux/fs.h> /* * 1: register the major number * 2: register callback functions for various capabilities * 3: register a request function * 4: disks characteristics information; gendisk */ #define RAMDK_MAJOR 166 //unique but static on my current machine as of now #define BLKDEV_NAME "ramdk" #define RAMDK_MINOR_NR 1 #define DISKSIZE 256*1024 #define NSECTORS 512 char buffer[DISKSIZE]; static struct gendisk *rdk = NULL; static DEFINE_SPINLOCK(ramdk_sp_lock); static struct request_queue *ramdk_queue = NULL; int ramdk_open(struct block_device *, fmode_t); int ramdk_release(struct gendisk *, fmode_t); int ramdk_open(struct block_device *blk, fmode_t mode) { printk(KERN_INFO "place holder for ramdisk's open method"); return 0; } int ramdk_release(struct gendisk *gdk, fmode_t mode) { printk(KERN_INFO "place holder for ramdisk's release method"); return 0; } static const struct block_device_operations ramdk_op = { .owner = THIS_MODULE, .open = ramdk_open, .release = ramdk_release, }; /* * block devices do not provide read()/write() routines like the char * devices, instead they use request callback. */ static void rdk_request(struct request_queue *q) { struct request *rq; /* * look at a request and then dequeue it */ rq = blk_fetch_request(q); while (rq) { unsigned long offset = blk_rq_pos(rq); unsigned long nbytes = blk_rq_cur_bytes(rq); int err = 0; while (nbytes) { if (rq_data_dir(rq) == READ) { memcpy(rq->buffer, (char *)offset, nbytes); } else if (rq_data_dir(rq) == WRITE) { memcpy((char *)offset, rq->buffer, nbytes); } else { printk(KERN_ERR "unknown operation\n"); } nbytes -= offset; } if (!__blk_end_request_cur(rq, err)) rq = blk_fetch_request(q); } return; } static int __init ramdk_init(void) { int ret = -1; /* * blocking call. On success, assign an unused major number and add a entry in * /proc/devices. */ if (register_blkdev(RAMDK_MAJOR, BLKDEV_NAME)) return -EBUSY; printk(KERN_INFO "registered block device %s with major: %d", BLKDEV_NAME, RAMDK_MAJOR); rdk = alloc_disk(RAMDK_MINOR_NR); if (!rdk) { ret = -ENOMEM; goto disk_alloc_fail; } rdk->fops = &ramdk_op; /* * HW perform I/O in the multiples of sectors(512Bytes, typically), whereas SW(FS, etc) * will work on block size(4k, typically). Therefore we need to tell the upper layers * about the capability of the hardware. This also sets the maximum number of sectors * that my hardware can receive per request. */ set_capacity(rdk, DISKSIZE*2); //Capacity, in terms of sectors /* * returns request queue for the block device. protected using spin lock */ ramdk_queue = blk_init_queue(rdk_request, &ramdk_sp_lock); if (!ramdk_queue) goto queue_fail; rdk->queue = ramdk_queue; rdk->major = RAMDK_MAJOR; rdk->first_minor = 0; sprintf(rdk->disk_name, BLKDEV_NAME); rdk->private_data = buffer; /* * Going live now!!! */ add_disk(rdk); return 0; queue_fail: printk(KERN_ERR "failed to allocate queue for %s",BLKDEV_NAME); disk_alloc_fail: unregister_blkdev(RAMDK_MAJOR, BLKDEV_NAME); return ret; } static void __exit ramdk_exit(void) { del_gendisk(rdk); put_disk(rdk); blk_cleanup_queue(ramdk_queue); unregister_blkdev(RAMDK_MAJOR, BLKDEV_NAME); printk(KERN_INFO "%s is offline now!!!",BLKDEV_NAME); } module_init(ramdk_init); module_exit(ramdk_exit); MODULE_LICENSE("GPL v2"); MODULE_AUTHOR("goon"); Once this issue is fixed, I plan to add support for filesystem related operations such as mkfs, mount, etc. [1] drivers/block/z2ram.c -Amit

3 3

hello and wiki edit
by Sven Dowideit May 1, 2013

May 1, 2013

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello there, I was browsing around and noticed that the wiki mentions the oftc #kernelnewbies group, which doesn't appear to be active at the moment. Assuming I'm correct, I'd like to update things like that - could I be added to the editor's list? Cheers Sven -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.12 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAlF/qFcACgkQPAwzu0QrW+mkUgCeLDzBUlsRd6RRzeLDe/6P/daO 1ysAn2c3GoNQwyh14k+zKzn7abfawgtq =qMQc -----END PGP SIGNATURE-----

4 5

Fwd: Need a good document that explains how to analyze kernel crash dump.
by amit mehta May 1, 2013

May 1, 2013

Forgot to reply-all On Wed, May 1, 2013 at 5:40 PM, Prasad Lakshman <prasad.bl.emb(a)gmail.com> wrote: > Thanks for your reply ,but the link you provided is explaining how to use > crash utility .I think it's tool by Redhat. But what I am looking for is > different. crash is an open source tool, you can have it on non rhel flavors as well. > I need to analyze when linux kernel crashes it generates a dmesg log ,i am > looking for document explains this dmesg crash log. > When your kernel crashes, those crash logs may not be available in the /var/log/messages (forget dmesg, it's a circular buffer that would be consumed even earlier). You need to configure something like kdump, netdump to have the kernel core dumped somewhere to be analysed later using tools such as crash. You may want to refer this [1], a very elaborate step by step guide on configuring kdump on your Linux machine. http://www.dedoimedo.com/computers/kdump.html - Amit

2 2

searchability of kernelnewbies mailing list archive.
by Anuz Pratap Singh Tomar May 1, 2013

May 1, 2013

Hi all, We do get a lot of good content in this mailing list and a lot of question get repeated quite often. I tried to search for few topics using various strings in google, it doesn't seem to pick up the right content from the archive. I have been following this list for a while so I am sure these topics were discussed for sure say about 3-4 months back. If I go to to archive there is no way of searching for contents or if there is it is not very prominently visible. Could this behaviour be fixed and make this mailing list more searchable either through google or directly? I have seen a few sites use google as default searching mechanism for the site, can this feature be implemented to newbies site? -- Thank you Warm Regards Anuz

1 0