On Thursday 17 March 2005 16:40, Tsillas, Demetrios J wrote: > I'm trying to simulate some hardware with a driver > which maps a region of kernel memory (allocated with > kmalloc) to user space. > > The user is called mmap which gives him direct access > to some register space. Are you using 2.4? On 2.6.11 remap_page_range does not exists, so I guess yes (on earlier 2.6 it existed though).
> This works well on a real Linux box but when I try > to simulate the register region with kernel memory, > the writes from user space do not go into the kmalloc'ed > region. Have you tried on a real Linux box with kmalloc'ed memory, just for the sake of it? Also, have you compared the code with http://www.xml.com/ldd/chapter/book/ch13.html#t1 (from Linux Device Drivers) and with drivers/char/mem.c: mmap_mem? I'm particularly suspicious about the use of VM_RESERVED > Here are the pointers: > > 0x27 1f0 000 is returned from kmalloc > 0x1f 1f0 000 is the above after a virt_to_phys I.e. after subtracting 0x08 000 000, i.e. 128 M, which should be the value of uml_physmem... a bit strange. Actually, it's normal in SKAS mode and impossible in TT one, so you're running in SKAS mode. In TT mode the value (set by arch/um/kernel/um_arch.c, in linux_main) should be 3G - 0.5G * (CONFIG_KERNEL_HALF_GIGS + CONFIG_NEST_LEVEL). In SKAS mode, instead, it should be anyway bigger: return(((unsigned long) set_task_sizes_skas) & ~0xffffff); That is a rounded pointer into the text section, and executables are mapped starting from 0x08048000, which is just above 128M... Ok, right, that value is correct, and you are running in SKAS mode. > 0x404e9000 is the user's virtual address > > I use remap_page_range with: > > 0x404e9000, 0x1f1f0000, 65536, 0x39 Ok, just verified it's correct... > The user writes to some offset in this region > (using his virtual address). Reads back the value, > all is good. > > The kernel tries to read the value using the > virtual address (offset from 0x271f0000) and gets > back zeros. I've also tried look at the phys addr > (offset from 0x1f1f0000) and that too is zeros. > > Any help appreciated. I've not looked very well to the above, and after I looked well it seems even less related, however the difference between i386 and Uml I can think of is: The first thing that comes to my mind is that while on i386, if you read datas from userspace without copy_from_user, it mostly works, until the page is swapped or unavailable; while on Uml SKAS mode (like i386 4G/4G mode) without copy_from_user you get meaningless datas (when it must be used); also one copy_from_user in excess is also dangerous in SKAS mode. -- Paolo Giarrusso, aka Blaisorblade Linux registered user n. 292729 http://www.user-mode-linux.org/~blaisorblade ------------------------------------------------------- SF email is sponsored by - The IT Product Guide Read honest & candid reviews on hundreds of IT Products from real users. Discover which products truly live up to the hype. Start reading now. http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click _______________________________________________ User-mode-linux-user mailing list User-mode-linux-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user
