[PATCH] Re: [uml-devel] mem= and large file support

["D. Bahi" <dbahi@enterasys.com>]

[-- Attachment #1.1: Type: text/plain, Size: 2435 bytes --]

yep that does it. here's the simple patch. and proof ;)

joy

linux umid=uml_1 mem=2176M root=/dev/ubd/0 \
ubd0=/local/umltest/cow/ubd0_1,/local/dbahi/root_fs.mv-31 \
ubd1=/local/umltest/cow/ubd1_1,/local/dbahi/swap_fs.256 \
uml_dir=/local/umltest/umldir hostfs=/local/umltest/hostfs/host_1 \
eth0=daemon,fe:fd:00:00:00:1,unix,/local/umltest/pipe

...

root@(none):~# cat /proc/meminfo
         total:    used:    free:  shared: buffers:  cached:
Mem:  2230177792 17244160 2212933632        0   466944  3371008
Swap:        0        0        0
MemTotal:      2177908 kB
MemFree:       2161068 kB
MemShared:           0 kB
Buffers:           456 kB
Cached:           3292 kB
SwapCached:          0 kB
Active:           2032 kB
Inactive:         2428 kB
HighTotal:           0 kB
HighFree:            0 kB
LowTotal:      2177908 kB
LowFree:       2161068 kB
SwapTotal:           0 kB
SwapFree:            0 kB

D. Bahi wrote:

> cool - so... it's in tempfile.c : 53
> make_tempfile calls mkstemp -
> 
> so maybe a build option required to get the
> option in with this library call? but which?
> 
> info on libc says:
> 
>      When the sources are compiled with `_FILE_OFFSET_BITS == 64' on a
>      32-bit system this function is in fact `tmpfile64', i.e. the LFS
>      interface transparently replaces the old interface.
> 
> so i'll try that.
> 
> and just for 'create_mem_file' completeness
> the devanon case doesn't need attention with
> the mmap. right?
> 
> -----
> 
> rt_sigprocmask(SIG_UNBLOCK, [ALRM VTALRM], [ALRM VTALRM IO], 8) = 0
> rt_sigprocmask(SIG_BLOCK, [IO], NULL, 8) = 0
> gettimeofday({1092914208, 599204}, NULL) = 0
> getpid()                                = 24405
> open("/tmp/vm_file-3NJodS", O_RDWR|O_CREAT|O_EXCL, 0600) = 3
> unlink("/tmp/vm_file-3NJodS")           = 0
> fchmod(3, 0777)                         = 0
> _llseek(3, 2147483648, [2147483648], SEEK_SET) = 0
> write(3, "\0", 1)                       = -1 EFBIG (File too large)
> --- SIGXFSZ (File size limit exceeded) @ 0 (0) ---
> +++ killed by SIGXFSZ +++
> 
> 
> Jeff Dike wrote:
> 
>> dbahi@enterasys.com said:
>>
>>> btw: with the 26-3um and 2.6.8.1 announce Jeff mentions the new SKAS,
>>> ! TT, STATIC allows mem= ~2.75G upper limit but do we need O_LARGEFILE
>>> somewhere? or to compile with -D_FILE_OFFSET_64_BITS 
>>
>>
>>
>> Probably.  Can you strace that and see where E2BIG is coming from?
>>
>>                 Jeff
>>
> 

-- 
db

[-- Attachment #1.2: gt2Mwohighmem.patch --]
[-- Type: text/x-diff, Size: 518 bytes --]

--- arch/um/Makefile@@/main/patch_kernel.org/patch_linux_kernel_uml_2.4.26/3	2004-08-13 22:22:48.000000000 -0400
+++ arch/um/Makefile	2004-08-22 22:34:08.000000000 -0400
@@ -51,7 +51,7 @@
 
 CFLAGS += $(ARCH_CFLAGS) $(CFLAGS-y) -D__arch_um__ -DSUBARCH=\"$(SUBARCH)\" \
 	-D_LARGEFILE64_SOURCE $(ARCH_INCLUDE) -Derrno=kernel_errno \
-	$(MODE_INCLUDE)
+	$(MODE_INCLUDE) -D_FILE_OFFSET_BITS=64
 
 check_gcc = $(shell if $(CC) $(1) -S -o /dev/null -xc /dev/null > /dev/null 2>&1; then echo "$(1)"; else echo "$(2)"; fi)
 

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 187 bytes --]