More flexible DMA allocations

2007-09-08 Thread Thomas Petazzoni
Hi, The report of Jonathan Corbet from the Kernel Summit at [1] contains the following sentences, without details: « The venerable DMA memory zone will go away, replaced by a more flexible way of allocating memory which meets specific requirements. » What's exactly going to be done ? I'm

Re: browsing kernel source with emacs

2007-09-08 Thread Pawel K
Are you generating the tags every time you search for it? I don't think so since generating cscope database takes about a minute on my machine while I wait about 10 secs for a tags list. Pawel Be a

Re: emacs xcscope.el slow with linux kernel

2007-09-08 Thread Mulyadi Santosa
Hi.. I can only answer this: 3. How should I generate cscope data files ? a. with cscope-indexer ? b. with make cscope directly from linux kernel ? c. with cscope -b -k -q ? back to 2.4 days, I use the (c) method, but now I simply just type make cscope when I am inside the kernel

Hard disk read-ahead hdparm

2007-09-08 Thread Ioannis Koltsidas
Developing a buffer manager for a database system, I want to find out how many pages the operating system fetches from the disk on every read operation (so that, on every read() I read into the bufferpool as many pages as possible). My questions are the following: a) Does the value of

request_irq() change?

2007-09-08 Thread Bob Beers
Where can I learn how to modify request_irq() call when migrating a driver from 2.6.15.x to 2.6.latest? I get this error: warning: passing arg 2 of `request_irq' from incompatible pointer type Thanks, -Bob -- To unsubscribe from this list: send an email with unsubscribe kernelnewbies to

Re: request_irq() change?

2007-09-08 Thread Thomas Petazzoni
Hi, Le Sat, 8 Sep 2007 13:59:25 -0400, Bob Beers [EMAIL PROTECTED] a écrit : Where can I learn how to modify request_irq() call when migrating a driver from 2.6.15.x to 2.6.latest? You can have a look at http://lwn.net/Articles/2.6-kernel-api/, which lists the API changes between all 2.6.x

Re: Hard disk read-ahead hdparm

2007-09-08 Thread Thomas Petazzoni
Hi, Le Sat, 8 Sep 2007 18:51:44 +0100, Ioannis Koltsidas [EMAIL PROTECTED] a écrit : a) Does the value of /sys/block/sda/queue/read_ahead_kb (typically 128k) decide the amount of data that every read() fetches from the disk ? Not for every read() : for uncached read()s only. E.g, if you

sys_call_ptr_t

2007-09-08 Thread Sachin Gaikwad
Hi all, Can somebody help me understand this ? typedef void (*sys_call_ptr_thttp://lxr.fsl.cs.sunysb.edu/linux/source/arch/x86_64/kernel/ident?i=sys_call_ptr_t)(void); What I am concluding is this - sys_call_table is an array to store address of functions and prototype of these functions is

New system call and User space header file

2007-09-08 Thread Sachin Gaikwad
Hi all, If I am adding a new system call sys_myroutine into the kernel then why is it necessary to have header file in user space like this - #include unistd.h _syscall2(int, myroutine, int arg1, int, arg2); What does it do when a user program calls this system call ? Can somebody explain this

Re: New system call and User space header file

2007-09-08 Thread Hareesh Nagarajan
On 9/8/07, Sachin Gaikwad [EMAIL PROTECTED] wrote: Hi all, If I am adding a new system call sys_myroutine into the kernel then why is it necessary to have header file in user space like this - #include unistd.h _syscall2(int, myroutine, int arg1, int, arg2); What does it do when a user

Re: sys_call_ptr_t

2007-09-08 Thread duryodhan
The sys call table is at arch/i386/kernel/syscall_table.S (for i386) it has stuff like .long sys_move_pages it doesn't have any sys_call_ptr_t. You don't need to assign a function. if you defined .long sys_move_pages then sys_move_pages is the name of the function that you will write as system

Re: New system call and User space header file

2007-09-08 Thread duryodhan
Hey, I would recommend that you directly call syscall yourself. #define __NR_syscall_number 324 //(e.g) #include linux/unistd.h #include sys/syscall.h in the C program do syscall(324, arg1, arg2) _syscall2 etc. is a pain to work with, and afaik has been unsupported for quite some time.