July 2011 - kernelnewbies@kernelnewbies.org - Kernelnewbies Mailing Lists

git format-patch: skipping interim commits.
by Ali Bahar July 24, 2011

July 24, 2011

Hi all, (yes, I _am_ aware that there is a separate git mailing list. This seemed to be more apt for _here_, as a first attempt.) When I run 'git format-patch' to prepare a submission, it creates a patch-file for every commit. While this is sometimes fine, it often isn't: I want it to skip all the interim commits which I did during development. There is no point in (and likely prosecutable in several jurisdictions) submitting the interim miscellany, thereby creating needless, non-Acked, log entries in master. I do have a crude work-around for this, but I prefer to find out how everyone else tackles this. I expect that it is more a _workflow_ issue than having to do with specifying revision ranges on the command-line. thanks, ali PS the crude work-around is to copy all the changes into a clean branch, and do a single commit. But that's error-prone when you have numerous files, and several sets of commits. PPS A simplified example, given a Master and a Dev branch: 1. Dev has lots of interim commits, for a single, logically-cohesive change. Once the work is complete, I want to create a single patch file for the entire delta against Master. 2. Dev will next have another set of commits. These are logically related and, once done, are to result in another single patch file. 3. and so on. This is all subject to Release Early, Release Often, so I am not talking about sitting on a bunch of changes for too long!

2 2

Memory change notifications
by Erlon Cruz July 24, 2011

July 24, 2011

Hi all, I have a device driver that implements a mmap in a char device. I would like to know if is possible to get notified(in that kernel module) every time a process change its mapped memory. Does anyone know if so? Regards, Erlon

5 5

How to recognize synchronous write requests
by Amirali Shambayati July 23, 2011

July 23, 2011

Hi all, As much as I know, type of an IO request can be realized by "DATA_DIR" macro,but I wanted to know if it is possible to distinguish a read or write is synchronous or asychronous. I'm also using "blktrace" to extract requests dispatching to disk. blktrace, determines synchrnous writes as "ws" and asynchronous writes as "w". Regards, -- Amirali Shambayati Bachelor Student Computer Engineering Department Sharif University of Technology Tehran, Iran

2 1

Problem installing vanilla kernel on Fedora 13
by Vaibhav Jain July 23, 2011

July 23, 2011

Hi, I am trying to install an extra vanilla kernel on Fedora 13. The Fedora 13 runs on vmware virtual machine. I installed it using and iso of live cd. A 'uname -r' gives the version as 2.6.33-85.fc13.i686. The kernel version I am trying to install is 2.6.38.8 I am following these steps : - Download kernel source and unzip it - make menuconfig (Saved the config) - make - make modules_install - make install However as soon as i do a 'make install ' an error message appears saying : "ERROR:modinfo:could not find module i2c_core" I could find the options for i2c in the config menu and enabled some of them but didn't help. Could anyone please tell me how to debug this error ? I have tried the same thing on Fedora 14 earlier and there was no such problem. Thanks Vaibhav Jain

2 2

GPIO acces on x86
by Victor Rodriguez July 23, 2011

July 23, 2011

Hi I want to access to the GPIO on a Intel Atom N450. I have already read the kernel GPIO documentation http://www.mjmwired.net/kernel/Documentation/gpio.txt And by sysfs (/sys/class/gpio) echo 19 > export shows me an error of permisions Do i have to recompile the kernel like with this instructions ? http://elinux.org/GPIO I really would prefer to make a char driver , I already have one for OMAP technology, on that case in order to access to gpio is necessary to do this : #define PIN_LED GPIO_TO_PIN(6, 12) #define MUX_LED DA850_GPIO6_12 /*Free gpio */ gpio_free(PIN_LED); /*Configure in mux.h */ ret = davinci_cfg_reg (MUX_LED); if (ret<0) { return ret; } /*Ask permision to turn it on */ ret = gpio_request(PIN_LED, "led"); if (ret) { return ret; } /*Configure as out put */ gpio_direction_output(PIN_LED, 1); /*TURN ON */ gpio_set_value(PIN_LED, 1); But how can I do the GPIO_TO_PIN(6, 12) on Atom technology ? Thanks a lot for your help Regards Sincerely yours Victor Rodriguez

1 0

mount FS without filesystem type
by piyush moghe July 22, 2011

July 22, 2011

I have a question regarding mounting filesystem without giving filesystem type as argument to mount command: AFAIK when mount command is run to mount a FS the corresponding call made in kernel is "compat_sys_mount" as per kernel 2.6.36. Now compat_sys_kernel expects a filesystem "type" name as argument but in compat_sys_mount there is no check for "type" argument as NULL, now as part of this mount path do_kern_mount also called which uses the "type" argument to compat_sys_mount and finds out the corresponding file_system_type structure from the file_system_type structure list and if not matching structure is found it returns with error. 842 asmlinkage long compat_sys_mount(const char __user * dev_name, 843 const char __user * dir_name, 844 const char __user * type, unsigned long flags, <====== type argument to compat_sys_mount 845 const void __user * data) 846 { 847 char *kernel_type; 848 unsigned long data_page; 849 char *kernel_dev; 850 char *dir_page; 851 int retval; 852 853 retval = copy_mount_string(type, &kernel_type); 1098 do_kern_mount(const char *fstype, int flags, const char *name, void *data) 1099 { 1100 struct file_system_type *type = get_fs_type(fstype); <======= same type argument to search in list 1101 struct vfsmount *mnt; 1102 if (!type) <======= if the type name entry not found in list return ERRROR 1103 return ERR_PTR(-ENODEV); 1104 mnt = vfs_kern_mount(type, flags, name, data); 1105 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && 1106 !mnt->mnt_sb->s_subtype) 1107 mnt = fs_set_subtype(mnt, fstype); 1108 put_filesystem(type); 1109 return mnt; 1110 } Now my question is that if I don't provide FS type name to mount command how the FS gets mounted. Is there some other path for this code or if before calling compat_sys_mount i.e in user mode FS type is determined somehow? Regards, Piyush

3 4

Problem installing vanilla kernel on Fedora 13
by srivatsa bhat July 22, 2011

July 22, 2011

Hi Vaibhav, >Hi, >I am trying to install an extra vanilla kernel on Fedora 13. The Fedora 13 runs on vmware virtual machine. >I installed it using and iso of live cd. A 'uname -r' gives the version as 2.6.33-85.fc13.i686. >The kernel version I am trying to install is 2.6.38.8 >I am following these steps : >- Download kernel source and unzip it >- make menuconfig (Saved the config) >- make >- make modules_install >- make install >However as soon as i do a 'make install ' an error message appears saying : >"ERROR:modinfo:could not find module i2c_core" >I could find the options for i2c in the config menu and enabled some of them but didn't help. >Could anyone please tell me how to debug this error ? I have tried the same thing on Fedora 14 earlier and there was >no such problem. >Thanks >Vaibhav Jain Try running make localmodconfig after running make menuconfig. The former command finds out all the currently loaded kernel modules, device drivers etc and ensures that all the config options necessary for them are set in the .config file. Regards, Srivatsa S. Bhat

3 2

How to generate PAUSE frame
by bill July 22, 2011

July 22, 2011

Hi, all I have a 10G NIC card with fiber interface, this NIC is drived by a private os with l2 rxonly forwarding function, and I want to test this NIC could sent out PAUSE frame. host client +--------+ +------------+ | | fiber | 10G NIC | | NIC | <===> | l2fwd | | | | rxonly | +--------+ +------------+ If host send packet faster than client could receive, client 10G NIC should send out PAUSE frame, and use "ethtool -a eth0" on host to show any PAUSE frame received. How could I do this without any faster NIC card than 10G? thanks bill

2 2

How to submit multiple related patches
by Zac Storer July 22, 2011

July 22, 2011

What is the best way to send multiple related patches? for example: A series of patches that fix indentation issues within a single src file.

2 1

How to write portable device drivers?
by gmane＠reliableembeddedsystems.com July 21, 2011

July 21, 2011

Hi, Please correct me, if I'm wrong, but that's my understanding. include/linux/byteorder/generic.h defines macros, which take both the CPU endianess and the endianess of the peripherial device you want to access into account. e.g. cpu_to_le32 will no matter what's the endianess of the CPU write to a little endian 32 address (do the swapping or not) but there are also in lib/iomap.c macros like iowrite32, which don't seem to take the CPU endianess into account Now looking e.g. at drivers/net/e100.c both kind of macros are used. e.g. ioread32 and also cpu_to_le32 The iowritexxx macros seem to boil down to outxx macros and it looks like iomap needs to be called before - maybe the endianess of the CPU is taken into account there, if so it's not obvious to me;) I'm a bit confused since we seem to have I/O ports and I/O memory. Can you please elaborate a bit when to use what and what happens with the CPU endianess in case of ioread/write macros? My goal is to understand how to write portable drivers which can be used with all kinds of funny CPU / peripheral endianess combinations. Regards, Robert ..."Everything should be built top-down, except the first time." My public pgp key is available at: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x90320BF1

1 0