> > >From what I read in the mailing list, the principle of "never doing > > this or that from inside the kernel" is repeated thousand times a day. > > We highly appreciate the tireless education of gurus in the list to > > the newbies like us ;-) > > > > For example, to read/write to file from kernelspace is always > > undesired. And there is a solution I got from the list: to use > > request_firmware() instead. > > > > But how to use request_firmware() for that trick (to read/write to > > file)? Is that encouraged/usual solution? I looked for some examples > > about the technique, but to no avail. > > request_firmware() is for a specific case of loading data that hardware > devices need to operate. > > For different things you want different solutions. > > If you just want to load or save a configuration, you probably want to > talk to a user-space process instead. You can use proc file, device or > netlink socket for that. You can even start the process with > call_usermodehelper() if there is no suitable process waiting on your > proc-file/device/netlink-socket. > > The reason is, that the user-mode helper can then implement various > policies without modifying the kernel. Eg. there might be no writable > filesystem, but the helper can send the data to a server. If you > insisted on writing to a file, it would be a problem. OK, let me show an example. I just supposed somehow I modify the ELF loader (by patching the kernel), and I want the new loader, after loading the executable into memory, to write the memory process to somewhere on the disk. How can I do this, if I cannot use read/write syscall from inside the kernel? I am sorry if the above example is silly. I just suppose it is the case to clarify my confusion ;-) Thank you a lot, AQ -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
