Messages in this thread | | | Date | Fri, 13 Nov 1998 22:52:32 +0100 (CET) | From | Cyrille Chepelov <> | Subject | [NEW PATCH] Re: (i386) using int15/E820 to get the sys memory map |
| |
(new version available : http://www.crans.ens-cachan.fr/~chepelov/linux/qsam-2.1.128.patch
* moved the 16-bit-to-32-bit sysmap buffer up a little to make room for 764 bytes' worth of command line, as suggested by Colin Plumb. * suppressed the __initfunc sysmemmap_show(), which was made obsolete by the /proc interface, as you pointed out.
Should apply clean at least on 2.1.12[78])
On Fri, 13 Nov 1998, Jeanette Pauline Middelink wrote:
> A few questions are left: > - Why did you call it QSAM? why not ACPI?
"Query System Address Map" -- Intel's official name for int15/E820 (as given by the ACPI 1.0a spec, and some bits on developer.intel.com)
> - Did you hear from another developers about ACPI? I'm currently Besides you and Philip Gladstone, I know of no-one active on that topic. But there might be already enough people to warrant a special forum (could be nice if there was something mildly ready for 2.3.2 lobbying <grin>)
> investigating to make the ACPI tables visible under /proc/acpi, > and it seems your sysmemmap could nicely integrate in that.
That's for the exact same reason I wrote the qsam patch :-) This is my working sketch, if you want to have a look at it : http://www.crans.ens-cachan.fr/~chepelov/linux/acpi-cc-19981113.tar.gz (a standalone module, really rough. In particular, it won't populate the /proc/acpi/tree until I advance a little the AML interpreter. The AML namespace management routines are a real mess : I'm really new to dentry wizardry... Thanks LXR ! I just wish there was an easy way to map from a struct proc_dir_entry to either a regular inode or a dentry (but my knoweldge of these things is really too parcellar. I plan to throw a lot of this code, anyway :-)
One thing it does, however, is to report a few things in /proc/acpi : rsd_ptr, rsdt, facs, facp, dsdt [I just found that I forgot to set /proc/acpi/. r-xr-xr-x instead of 666. bwah.])
> - Indeed the AML to <whatever> is a great ball, did those designers > really think BIOSes have AML interpreters like that?? :)
Well, the BIOS doesn't have an interpreter, the OS is supposed to. And you've seen who signed that spec :-)
> - That _initfunc is not really needed, isn't it? beter remove it > than, since /proc/sysmemmap will be sufficient...
Yes, indeed I didn't clean things up sufficiently.
> System memory map entries: 8 > Addr=0x0000000000000000 Length=0x00000000000a0000 Type=1 AddressRangeMemory > Addr=0x00000000000f0000 Length=0x0000000000010000 Type=2 AddressRangeReserved > Addr=0x0000000000100000 Length=0x0000000007efd000 Type=1 AddressRangeMemory > Addr=0x0000000007ffd000 Length=0x0000000000002000 Type=3 AddressRangeACPI > Addr=0x0000000007fff000 Length=0x0000000000001000 Type=4 AddressRangeNVS > Addr=0x00000000fec00000 Length=0x0000000000001000 Type=2 AddressRangeReserved > Addr=0x00000000fee00000 Length=0x0000000000001000 Type=2 AddressRangeReserved > Addr=0x00000000ffff0000 Length=0x0000000000010000 Type=2 AddressRangeReserved
Hmmmm... it looks like a common practice from BIOS people to put the ACPI tables at the end. That might even help some sparse-memory impaired OSes <grin> (it might simply be an ASUS/Award practice. I've seen sysmemmaps only from those boards).
-- Cyrille
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |