October 2009 - openfirmware - mail.coreboot.org

r1402 - dev/olpc/dcon
by svn＠openfirmware.info Oct. 7, 2009

Oct. 7, 2009

Author: wmb Date: 2009-10-08 07:13:30 +0200 (Thu, 08 Oct 2009) New Revision: 1402 Modified: dev/olpc/dcon/dcon.fth Log: OLPC - Added backlight-{on,off} to DCON driver for XO-1. Modified: dev/olpc/dcon/dcon.fth =================================================================== --- dev/olpc/dcon/dcon.fth 2009-10-07 18:56:31 UTC (rev 1401) +++ dev/olpc/dcon/dcon.fth 2009-10-08 05:13:30 UTC (rev 1402) @@ -121,6 +121,9 @@ : brighter ( -- ) bright@ 1+ h# f min bright! ; : dimmer ( -- ) bright@ 1- 0 max bright! ; +: backlight-off ( -- ) mode@ 8 invert and mode! ; +: backlight-on ( -- ) mode@ 8 or mode! ; + \ Colour swizzle, AA, no passthrough, backlight : set-color ( color? -- ) dup to color?

1 0

r1401 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Oct. 7, 2009

Oct. 7, 2009

Author: wmb Date: 2009-10-07 20:56:31 +0200 (Wed, 07 Oct 2009) New Revision: 1401 Modified: cpu/x86/pc/olpc/via/devices.fth cpu/x86/pc/olpc/via/usb.fth Log: OLPC trac 9480 - fixed net devalias - was /usb/wlan, now /wlan for SD-connected wlan module on XO-1.5 Modified: cpu/x86/pc/olpc/via/devices.fth =================================================================== --- cpu/x86/pc/olpc/via/devices.fth 2009-10-07 08:52:49 UTC (rev 1400) +++ cpu/x86/pc/olpc/via/devices.fth 2009-10-07 18:56:31 UTC (rev 1401) @@ -337,6 +337,7 @@ devalias ide /ide@0/disk devalias sd /sd/disk@2 devalias ext /sd/disk@2 +devalias net /wlan \ The "int" devalias is defined in report-disk at runtime, since it \ varies between A-test and later boards. @@ -376,7 +377,7 @@ \ If there is a USB ethernet adapter, use it as the default net device. \ We can't use ?report-device here because we already have net aliased -\ to /usb/wlan, and ?report-device won't override an existing alias. +\ to /wlan, and ?report-device won't override an existing alias. : report-net ( -- ) " /usb/ethernet" 2dup locate-device 0= if ( name$ phandle ) drop ( name$ ) Modified: cpu/x86/pc/olpc/via/usb.fth =================================================================== --- cpu/x86/pc/olpc/via/usb.fth 2009-10-07 08:52:49 UTC (rev 1400) +++ cpu/x86/pc/olpc/via/usb.fth 2009-10-07 18:56:31 UTC (rev 1401) @@ -41,7 +41,6 @@ pwd$ $nopage-show-devs then ; -devalias net /usb/wlan : probe-usb ( -- ) ." USB2 devices:" cr

1 0

r1400 - ofw/fs/ext2fs
by svn＠openfirmware.info Oct. 7, 2009

Oct. 7, 2009

Author: wmb Date: 2009-10-07 10:52:49 +0200 (Wed, 07 Oct 2009) New Revision: 1400 Modified: ofw/fs/ext2fs/dir.fth ofw/fs/ext2fs/layout.fth ofw/fs/ext2fs/methods.fth Log: OLPC trac 6210 - fixed rmdir and mkdir for ext2 file systems. Modified: ofw/fs/ext2fs/dir.fth =================================================================== --- ofw/fs/ext2fs/dir.fth 2009-10-06 23:56:37 UTC (rev 1399) +++ ofw/fs/ext2fs/dir.fth 2009-10-07 08:52:49 UTC (rev 1400) @@ -3,8 +3,9 @@ decimal -2 constant root-dir# +2 constant root-dir# 0 instance value dir-block# +0 instance value lblk# variable diroff variable totoff @@ -74,16 +75,16 @@ \ After this point, the code should be independent of the disk format! : init-inode ( mode inode# -- ) - inode >r ( mode ) - r@ /inode erase ( mode ) - r@ short! ( ) - time&date >unix-seconds ( time ) - dup r@ 2 la+ int! ( time ) \ set access time - dup r@ 3 la+ int! ( time ) \ set creation time - r@ 4 la+ int! ( ) \ set modification time - 1 r@ d# 13 wa+ short! \ set links_count to 1 - update - r> drop + inode >r ( mode r: inode-adr ) + r@ /inode erase ( mode r: inode-adr ) + r@ short! ( r: inode-adr ) + time&date >unix-seconds ( time r: inode-adr ) + dup r@ 2 la+ int! ( time r: inode-adr ) \ set access time + dup r@ 3 la+ int! ( time r: inode-adr ) \ set creation time + r@ 4 la+ int! ( r: inode-adr ) \ set modification time + 1 r@ d# 13 wa+ short! ( r: inode-adr ) \ set links_count to 1 + update ( r: inode-adr ) + r> drop ( ) ; \ On entry: @@ -147,7 +148,7 @@ : to-previous-dirent ( -- ) diroff @ ( this ) - 0 diroff ! ( this ) + diroff off ( this ) begin ( this ) dup diroff @ dirent-len@ + <> ( this not-found? ) while ( this ) @@ -178,17 +179,19 @@ 7 constant symlink-type : ($create) ( name$ mode -- error? ) - >r ( name$ ) + >r ( name$ r: mode ) \ check for room in the directory, and expand it if necessary - dup >reclen no-dir-space? if ( name$ new-reclen ) + dup >reclen no-dir-space? if ( name$ new-reclen r: mode ) \ doesn't fit, allocate more room - bsize ( name$ bsize ) + bsize ( name$ bsize r: mode ) append-block lblk#++ get-dirblk drop - then ( name$ rec-len ) + then ( name$ rec-len r: mode ) \ At this point dirent points to the place for the new dirent - regular-type alloc-inode set-dirent false ( error? ) + \ XXX handle symlinks! + r@ h# f000 and h# 4000 = if dir-type else regular-type then ( name$ rec-len type r: mode ) + alloc-inode set-dirent false ( error? r: mode ) r> dirent-inode@ init-inode ; @@ -236,7 +239,7 @@ : first-dirent ( dir-inode# -- end? ) \ Adapted from (init-dir) set-inode get-dirblk if true exit then - 0 diroff ! 0 totoff ! ( ) + diroff off totoff off ( ) false ( ) ; @@ -347,6 +350,16 @@ wd-inum first-dirent ; +\ Returns true if inode# refers to a directory that is empty +\ Side effect - changes dirent context +: empty-dir? ( -- empty-dir? ) + dir? 0= if false exit then + + file-handle first-dirent if false exit then \ Should be pointing to "." entry + next-dirent if false exit then \ Should be point to ".." entry + next-dirent ( end? ) \ The rest should be empty +; + \ Delete a file, given its inode. Does not affect the directory entry, if any. : idelete ( inode# -- ) to inode# @@ -366,7 +379,7 @@ inode# >r file-handle set-inode - file? 0= if r> drop true exit then \ XXX handle symlinks + file? 0= if r> drop true exit then \ XXX handle symlinks \ is this the only link? link-count dup 2 u< if Modified: ofw/fs/ext2fs/layout.fth =================================================================== --- ofw/fs/ext2fs/layout.fth 2009-10-06 23:56:37 UTC (rev 1399) +++ ofw/fs/ext2fs/layout.fth 2009-10-07 08:52:49 UTC (rev 1400) @@ -7,7 +7,6 @@ 0 instance value #ind-blocks1 0 instance value #ind-blocks2 0 instance value #ind-blocks3 -0 instance value lblk# \ first number \ 0 12 direct Modified: ofw/fs/ext2fs/methods.fth =================================================================== --- ofw/fs/ext2fs/methods.fth 2009-10-06 23:56:37 UTC (rev 1399) +++ ofw/fs/ext2fs/methods.fth 2009-10-07 08:52:49 UTC (rev 1400) @@ -12,11 +12,14 @@ : $create ( name$ -- error? ) o# 100666 ($create) ; -\ XXX note: increment link count in parent + : $mkdir ( name$ -- error? ) o# 40777 ($create) if true exit then - + link-count 1+ link-count! \ The ".." entry is another link to parent + file-handle to inode# + link-count 1+ link-count! \ The "." entry is another link to the new directory + add-block ( block# ) file-size h# 400 + file-size! dup direct0 int! update ( block# ) @@ -37,27 +40,26 @@ ; : $delete! $delete ; \ XXX should these be different? -\ XXX note: decrement link count in parent -: $rmdir ( name$ -- error? ) \ XXX UNTESTED +: $rmdir ( name$ -- error? ) $find-file if true exit then ( ) wf-type dir-type <> if true exit then ( ) + \ Now the dirent is the one for the directory to delete and the + \ inode is for the parent directory - inode# >r \ save parent directory - file-handle set-inode - dir? 0= if r> drop true exit then - - (delete-files) file-handle if r> drop true exit then \ still some left - - \ now empty, remove it. - delete-blocks + dirent-vars 2>r 2>r \ save lookup state in parent directory + file-handle set-inode \ Switch the inode to the directory to delete - \ delete inode. clear or mark it? - file-handle free-inode + empty-dir? 0= if 2r> 2r> 4drop true exit then + + \ The directory is empty so it's okay to remove it + delete-blocks \ Release the data blocks used for dirent storage + inode# free-inode \ Release the inode number - r> to inode# \ restore parent directory + 2r> 2r> restore-dirent \ restore lookup state in parent directory - \ delete directory entry - del-dirent ( error? ) + del-dirent if true exit then \ Remove the dirent from the parent directory + link-count 1- link-count! \ Decrement parent link count + false ; headers

1 0

r1399 - cpu/x86/pc/olpc/via ofw/fs/ext2fs
by svn＠openfirmware.info Oct. 6, 2009

Oct. 6, 2009

Author: wmb Date: 2009-10-07 01:56:37 +0200 (Wed, 07 Oct 2009) New Revision: 1399 Modified: cpu/x86/pc/olpc/via/fw.bth ofw/fs/ext2fs/dir.fth ofw/fs/ext2fs/methods.fth Log: OLPC trac 9475 - emacs was crashing when you do a "save-file" due to an error in the ext2fs implementation of $delete! . Modified: cpu/x86/pc/olpc/via/fw.bth =================================================================== --- cpu/x86/pc/olpc/via/fw.bth 2009-10-06 21:49:09 UTC (rev 1398) +++ cpu/x86/pc/olpc/via/fw.bth 2009-10-06 23:56:37 UTC (rev 1399) @@ -300,6 +300,9 @@ fload ${BP}/ofw/gui/loadmenu.fth fload ${BP}/ofw/gui/insticon.fth +\ Uninstall the diag menu from the general user interface vector +\ so exiting from emacs doesn't invoke the diag menu. +' quit to user-interface fload ${BP}/cpu/x86/pc/olpc/via/mfgtest.fth : screen-#lines ( -- n ) Modified: ofw/fs/ext2fs/dir.fth =================================================================== --- ofw/fs/ext2fs/dir.fth 2009-10-06 21:49:09 UTC (rev 1398) +++ ofw/fs/ext2fs/dir.fth 2009-10-06 23:56:37 UTC (rev 1399) @@ -365,7 +365,7 @@ file-handle 0= if true exit then inode# >r - file-handle set-inode if r> drop true exit then + file-handle set-inode file? 0= if r> drop true exit then \ XXX handle symlinks \ is this the only link? Modified: ofw/fs/ext2fs/methods.fth =================================================================== --- ofw/fs/ext2fs/methods.fth 2009-10-06 21:49:09 UTC (rev 1398) +++ ofw/fs/ext2fs/methods.fth 2009-10-06 23:56:37 UTC (rev 1399) @@ -43,7 +43,7 @@ wf-type dir-type <> if true exit then ( ) inode# >r \ save parent directory - file-handle set-inode if r> drop true exit then + file-handle set-inode dir? 0= if r> drop true exit then (delete-files) file-handle if r> drop true exit then \ still some left

1 0

r1398 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Oct. 6, 2009

Oct. 6, 2009

Author: wmb Date: 2009-10-06 23:49:09 +0200 (Tue, 06 Oct 2009) New Revision: 1398 Modified: cpu/x86/pc/olpc/via/resume.bth Log: Via resume.bth - eliminated a lot of ifdef'ed out cruft that was leftover from the Geode code and will never be used for the Via case. Modified: cpu/x86/pc/olpc/via/resume.bth =================================================================== --- cpu/x86/pc/olpc/via/resume.bth 2009-10-06 21:36:16 UTC (rev 1397) +++ cpu/x86/pc/olpc/via/resume.bth 2009-10-06 21:49:09 UTC (rev 1398) @@ -21,11 +21,6 @@ \ This code will be copied to RAM at suspend-base \ create measure-suspend -\ create save-display -\ create reset-smbus -\ create reset-smbus-bitbang -\ create restore-usb-power -\ create checksum-test : resume-progress ( byte -- ) " h# 34 # al mov al h# 70 # out ( byte ) # al mov al h# 71 # out" eval @@ -168,42 +163,20 @@ \ This is where Geode saves the PIT state \ This is where Geode saves the SMBUS controller state \ This is where Geode saves the MFGPT state -\ This is where Geode saves the CaFe configuration \ Other state we might want to save: \ ISA DMA Controller - \ AC97 + \ HDaudio \ Codec \ USB \ This is where Geode stops the video refresh -[ifdef] checksum-test - \ Checksum memory from 1M to top (excluding framebuffer) - h# 0010.0000 # si mov - h# 4d534b43 # ax mov ax stos \ Marker 'CKSM' - di resume-data h# 10 - #) mov \ Save checksum base address - begin - bx bx xor - h# 10.0000 2 rshift # cx mov \ Word count for 1MB - begin ax lods ax bx add loopa - ax stos - h# ec0.0000 # si cmp - = until -[then] - [ifdef] measure-suspend di 0 #) mov rdtsc ax h# 18 #) mov dx h# 1c #) mov [then] - \ This is where the Geode sets up the sleep/wakeup sequence timing - -[ifdef] notdef - \ Clear existing wakeup status - h# 400 port-rw op: ax dx out -[then] \ notdef - \ The caller chooses the wakeup conditions, so we don't do it here h# ffff h# 420 port-ww \ Clear all status bits @@ -215,19 +188,6 @@ wbinvd \ Flush the cache -[ifdef] delete-me - \ Setup the register values in advance so the active instruction sequence - \ is as short as possible, thus keeping all the activity in one cache line. -\ h# 0400 h# 404 port-ww \ Setup sleep type -[then] - -[ifdef] notdef - \ Align to a cache line boundary - also forth - begin here asm-base - h# 1f and while h# 90 c, repeat \ Align with NOPs - previous -[then] - h# 2400 h# 404 port-ww \ Go to sleep h# 1000000 # cx mov begin nop loopa \ Spin in this cache line while going down @@ -246,8 +206,6 @@ \ Running from a 32-bit identity-mapped code segment \ Using physical addresses -\ char < 3f8 port-wb begin 3fd port-rb 20 bitand 0<> until - resume-data # bp mov cld @@ -260,134 +218,8 @@ h# 21 resume-progress -[ifdef] checksum-test - \ Checksum memory from 1M to top (excluding framebuffer) - h# 0010.0000 # si mov - resume-data h# 10 - #) di mov \ Save checksum base address - begin - bx bx xor - h# 10.0000 2 rshift # cx mov \ Word count for 1MB - begin ax lods ax bx add loopa - ax 0 [di] cmp - <> if - char C 3f8 port-wb begin 3fd port-rb 20 bitand 0<> until - ret - then - 4 [di] di lea - h# ec0.0000 # si cmp - = until -[then] - h# 38 [bp] si lea \ Save area - h# 22 resume-progress - -[ifdef] reset-smbus-bitbang - \ GPIO15 is SMB_DATA - \ GPIO14 is SMB_CLOCK - h# 5140.000c rmsr ax bx mov - h# c000 # ax mov \ Mask to set SMB_DATA and SMB_CLOCK - h# 00 [bx] dx lea ax dx out \ Set output values to high - h# 04 [bx] dx lea ax dx out \ Set pins to output - h# c0000000 # ax mov \ Mask to clear SMB_DATA and SMB_CLOCK - h# 10 [bx] dx lea ax dx out \ Deselect OUT AUX1 - h# 14 [bx] dx lea ax dx out \ Deselect OUT AUX2 - h# 34 [bx] dx lea ax dx out \ Deselect IN AUX1 - - d# 16 # cx mov forget-msr \ Generate 8 low pulses on SMB_CLOCK - begin - \ 5 uS delay (slightly longer for GX) - rdtsc ax bx mov d# 5 d# 500 * # bx add - begin rdtsc bx ax sub 0>= until - - cx bx mov \ Save cx for use by rmsr - h# 5140.000c rmsr ax dx mov \ GPIO output register - h# 40000000 # ax mov ax dx out \ Clear SMB_CLOCK - bx cx mov forget-ms \ Restore cx - - \ 5 uS delay (slightly longer for GX) - rdtsc ax bx mov d# 5 d# 500 * # bx add - begin rdtsc bx ax sub 0>= until - - cx bx mov \ Save cx for use by rmsr - h# 5140.000c rmsr ax dx mov \ GPIO output register - h# 4000 # ax mov ax dx out \ Set SMB_CLOCK - bx cx mov forget-msr \ Restore cx - loopa - - \ 5 uS delay (slightly longer for GX) - rdtsc ax bx mov d# 5 d# 500 * # bx add - begin rdtsc bx ax sub 0>= until -[then] - - h# 23 resume-progress - -\ GPIO restore - - h# 24 resume-progress - -\ Display restore - -\ Flat panel turn-on - - h# 25 resume-progress - -\ PIC restore - - h# 26 resume-progress - -\ PIT restore - - h# 27 resume-progress - -\ SMBUS controller restore - -[ifdef] reset-smbus - \ This little dance resets the DCON's SMbus interface - \ We start with dx pointing to SMBUS reg 5 - dx dec dx dec \ Point back to smbus reg3 (control 1) - h# 1 # al mov \ START - al dx out \ Initiate an SMBUS cycle to the DCON - - \ We could split here and move the following down, so as to overlap - \ the delay time with other work, but it doesn't matter because the - \ later CaFe chip setup will stall anyway. - - d# 32 # cx mov forget-msr \ Loop count (usually ready in 20 uS) - dx dec dx dec \ SMBUS reg1 (status) - begin - dx al in - h# 40 # al test - loope \ Wait for ready to accept byte or timeout - - h# 1a # al mov \ Address byte - dx dec \ SMBUS reg0 (data) - al dx out \ Initiate address out cycle - - \ Another possible split point, in case we should need to overlap - d# 256 # cx mov forget-msr \ Loop count (usually ready in 172 uS) - dx inc \ SMBUS reg1 (status) - begin - dx al in - h# 50 # al test - loope \ Wait for done or error or timeout - - h# 2 # al mov \ Stop - dx inc dx inc \ SMBUS reg3 (control 1) - al dx out \ Initiate STOP - - h# 10 # al mov \ Ack NEGACK - dx dec dx dec \ SMBUS reg1 (status) - al dx out \ While acking the NEGACK - \ End of DCON SMbus reset dance -[then] - - h# 28 resume-progress - -\ MFGPTs restore - - h# 29 resume-progress - \ D12F0 SDIO h# 6010 config-setup ax lods ax dx out \ BAR0 h# 6014 config-setup ax lods ax dx out \ BAR1 @@ -448,15 +280,6 @@ h# 2a resume-progress -\ Restore CaFe configuration - - h# 2b resume-progress - -\ Display base addresses - - \ There is a lot of other stuff that must be done to turn on the - \ video - but we will let the gx driver take care of that. - \ XXX keyboard init \ DCON fiddling \ USB @@ -501,8 +324,6 @@ \ Identity mapping of low memory might not exist now -\ char > 3f8 port-wb begin 3fd port-rb 20 bitand 0<> until - gs pop fs pop

1 0

r1397 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Oct. 6, 2009

Oct. 6, 2009

Author: wmb Date: 2009-10-06 23:36:16 +0200 (Tue, 06 Oct 2009) New Revision: 1397 Modified: cpu/x86/pc/olpc/via/acpi.fth cpu/x86/pc/olpc/via/addrs.fth cpu/x86/pc/olpc/via/resume.bth cpu/x86/pc/olpc/via/romreset.bth Log: Via - implement X_firmware_waking_vector support in early startup code, and use it in the Forth resume code to eliminate hardcoded addresses. Modified: cpu/x86/pc/olpc/via/acpi.fth =================================================================== --- cpu/x86/pc/olpc/via/acpi.fth 2009-10-06 11:07:53 UTC (rev 1396) +++ cpu/x86/pc/olpc/via/acpi.fth 2009-10-06 21:36:16 UTC (rev 1397) @@ -249,20 +249,20 @@ 0 w, 0 w, \ Room for the segment:offset pointer then - op: ax bx mov - h# 0f # ax and ax cx mov - op: 4 # bx shr + op: ax bx mov \ Get linear address to jump to - must be < 1M + h# 0f # ax and ax cx mov \ CX: low 4 bits for offset of seg:off address + op: 4 # bx shr \ BX: high 16 bits for segment of seg:off \ Set data segment for storing offset and segment below ax ax xor ds16 # al mov ax ds mov \ 16-bit data segment - cx wake-adr wa1+ #) mov \ Offset - bx wake-adr la1+ #) mov \ Segment + cx wake-adr wa1+ #) mov \ Put offset at wake-adr+2 + bx wake-adr la1+ #) mov \ Put segment at wake-adr+4 cr0 ax mov h# fe # al and ax cr0 mov \ Enter real mode here 5 + do-acpi-wake - wake-adr + lwsplit d# 12 lshift #) far jmp \ Jump to set cs - cs: wake-adr wa1+ lwsplit drop s#) far jmp + cs: wake-adr wa1+ lwsplit drop s#) far jmp \ Jump to seg:off address stored at wake-adr+2 end-code here do-acpi-wake - constant /do-acpi-wake Modified: cpu/x86/pc/olpc/via/addrs.fth =================================================================== --- cpu/x86/pc/olpc/via/addrs.fth 2009-10-06 11:07:53 UTC (rev 1396) +++ cpu/x86/pc/olpc/via/addrs.fth 2009-10-06 21:36:16 UTC (rev 1397) @@ -43,7 +43,6 @@ 0 value dma-base \ Set in probemem.fth h# f.0000 constant suspend-base \ In the DOS hole -h# f.0008 constant resume-entry h# f.1000 constant resume-data \ If you change these, also change {g/l}xmsrs.fth and {g/l}xearly.fth Modified: cpu/x86/pc/olpc/via/resume.bth =================================================================== --- cpu/x86/pc/olpc/via/resume.bth 2009-10-06 11:07:53 UTC (rev 1396) +++ cpu/x86/pc/olpc/via/resume.bth 2009-10-06 21:36:16 UTC (rev 1397) @@ -32,6 +32,7 @@ ; start-assembling \ Turn on the target assembler +suspend-base to asm-origin \ This code will be copied to a fixed location protected-mode \ Suspend code - called as a subroutine, with return address on the stack @@ -41,15 +42,6 @@ \ is used as a memory pointer before and after the paging turn-off. label suspend-entry \ h# f.0000 - entry point at fixed location, called by OS - here h# 10 + #) jmp \ Jump to offset 10 - 8 pad-to - -label resume-entry \ h# f.0008 - entry point at fixed location, called by resume - e9 c, 0 , \ To be patched later - nop \ Aligns to even, for the benefit of the disassembler - h# 10 pad-to - -label suspend-save \ h# f.0010 (Page directory VA in ax) pusha pushf cli ds push @@ -66,9 +58,10 @@ 0 [si] cx mov forget-msr \ Save old PDIR entry 0 h# 83 # 0 [ax] mov \ Punch a 4M mapping 0->0 into the page directory cr3 ax mov ax cr3 mov \ Invalidate the TLB to activate the mapping - h# f0060 # ax mov ax jmp \ Jump to suspend-physical, disabling paging - h# 60 pad-to + \ Jump to the physical address of the next instruction, disabling paging + here 7 + asm-base - asm-origin + # ax mov ax jmp + \ We have to be running from a virtual=physical mapping here, otherwise we \ die on the instruction after paging is turned off. cr0 ax mov ax bx mov \ CR0 saved in BX @@ -93,9 +86,9 @@ h# a [bp] sidt \ Interrupt descriptor table register h# 10 [bp] sgdt \ Global descriptor table register - bx h# 18 [bp] mov - cr3 ax mov ax h# 1c [bp] mov - cr4 ax mov ax h# 20 [bp] mov + bx h# 18 [bp] mov \ CR0 + cr3 ax mov ax h# 1c [bp] mov \ CR3 + cr4 ax mov ax h# 20 [bp] mov \ CR4 cx h# 24 [bp] mov \ PDIR entry 0 si h# 28 [bp] mov \ PDIR VA di h# 2c [bp] mov \ Address offset @@ -216,6 +209,10 @@ h# fe resume-progress + \ Set the resume address +label resume-adr-loc + h# 0 # facs-adr h# 18 + #) mov \ Value will be patched below + wbinvd \ Flush the cache [ifdef] delete-me @@ -242,7 +239,7 @@ h# 3f # al mov al h# 70 # out h# 71 # al in al inc al h# 71 # out \ Resume code -here resume-entry put-branch +here asm-base - asm-origin + resume-adr-loc 6 + ! \ Assumptions: \ Processor is in protected mode @@ -474,9 +471,9 @@ h# 2c [bp] di mov \ VA of suspend-base in di h# 28 [bp] si mov \ PDIR VA h# 24 [bp] cx mov forget-msr \ PDIR entry 0 - h# 20 [bp] ax mov ax cr4 mov - h# 1c [bp] ax mov ax cr3 mov - h# 18 [bp] ax mov + h# 20 [bp] ax mov ax cr4 mov \ Restore CR4 + h# 1c [bp] ax mov ax cr3 mov \ Restore CR3 + h# 18 [bp] ax mov \ Get CR0 for a few lines later h# 10 [bp] lgdt \ Global descriptor table register h# a [bp] lidt \ Interrupt descriptor table register Modified: cpu/x86/pc/olpc/via/romreset.bth =================================================================== --- cpu/x86/pc/olpc/via/romreset.bth 2009-10-06 11:07:53 UTC (rev 1396) +++ cpu/x86/pc/olpc/via/romreset.bth 2009-10-06 21:36:16 UTC (rev 1397) @@ -227,15 +227,16 @@ cli + facs-adr h# 18 + #) ax mov ax ax or 0<> if \ X_Firmware_Waking_Vector + ax jmp \ Jump in 32-bit protected mode + then + facs-adr h# c + #) ax mov ax ax or 0<> if \ Resume in real mode for ACPI operating systems \ Linear wakeup address is in EAX wake-adr lwsplit drop cs16 #) far jmp - else - \ Resume in 32-bit protected mode for OFW - resume-data # sp mov - resume-entry # ax mov ax call \ This might return if checksumming fails then + char x report then \ Not a wakeup from S3

1 0

r1396 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Oct. 6, 2009

Oct. 6, 2009

Author: wmb Date: 2009-10-06 13:07:53 +0200 (Tue, 06 Oct 2009) New Revision: 1396 Modified: cpu/x86/pc/olpc/via/startcominit.fth Log: Via - poll the PSON gating state bit instead of using a time delay to determine when CMOS reads will work. Modified: cpu/x86/pc/olpc/via/startcominit.fth =================================================================== --- cpu/x86/pc/olpc/via/startcominit.fth 2009-10-06 03:28:24 UTC (rev 1395) +++ cpu/x86/pc/olpc/via/startcominit.fth 2009-10-06 11:07:53 UTC (rev 1396) @@ -29,9 +29,6 @@ \ First we check for a cached board ID in CMOS RAM, to avoid the \ possibly time-consuming operation of asking the EC. - d# 17 0 devfunc - end-table - \ Configure the I/O decoding to enable access to the EC \ Do this outside the if..then so the setup is consistent in all cases d# 17 0 devfunc @@ -55,13 +52,10 @@ 70 fb 82 mreg \ CPU to PCI flow control - CPU to PCI posted write, Enable Delay Transaction end-table - acpi-io-base 4 + port-rw \ Get APCI Status register - d# 10 # ax shr 7 # ax and 1 # ax cmp <> if \ Type 1 is wakeup from S3 - \ The following applies only to power-up; no delay is needed for wakeup from S3 - \ This delay is empirically necessary before reading CMOS - minimum is 36000 - about 50 ms - \ Before the delay has elapsed, the CMOS RAM returns 0 instead of the stored value. - d# 40000 wait-us - then + \ Wait until RTC PSON Gating is complete. See PG_VX855_VX875_092 page 139 (pdf page 160) + \ This takes about 48 mS in the power-on case, and is almost instantaneous in the + \ resume-from-S3 case due to the clearing of D17F0 Rx81[2] above. + begin 8882 config-rb h# 40 # al and 0<> until \ As an optimization to avoid long waits for the EC to respond, read the board ID \ that is cached in CMOS RAM.

1 0

r1395 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Oct. 5, 2009

Oct. 5, 2009

Author: wmb Date: 2009-10-06 05:28:24 +0200 (Tue, 06 Oct 2009) New Revision: 1395 Modified: cpu/x86/pc/olpc/via/versions.fth Log: Use new EC code version 1.9.10. Modified: cpu/x86/pc/olpc/via/versions.fth =================================================================== --- cpu/x86/pc/olpc/via/versions.fth 2009-10-06 03:26:47 UTC (rev 1394) +++ cpu/x86/pc/olpc/via/versions.fth 2009-10-06 03:28:24 UTC (rev 1395) @@ -2,10 +2,10 @@ \ The overall firmware revision macro: FW_MAJOR A -macro: FW_MINOR 11d +macro: FW_MINOR 11e \ The EC microcode -macro: EC_VERSION 1_9_9 +macro: EC_VERSION 1_9_10 \ Alternate command for getting EC microcode, for testing new versions. \ Temporarily uncomment the line and modify the path as necessary

1 0

r1394 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Oct. 5, 2009

Oct. 5, 2009

Author: wmb Date: 2009-10-06 05:26:47 +0200 (Tue, 06 Oct 2009) New Revision: 1394 Modified: cpu/x86/pc/olpc/via/romreset.bth cpu/x86/pc/olpc/via/startcominit.fth Log: OLPC trac 9451 - eliminated the startup delay before accessing CMOS RAM in the wakeup-from-S3 case, thus speeding-up resume. Modified: cpu/x86/pc/olpc/via/romreset.bth =================================================================== --- cpu/x86/pc/olpc/via/romreset.bth 2009-10-06 03:18:14 UTC (rev 1393) +++ cpu/x86/pc/olpc/via/romreset.bth 2009-10-06 03:26:47 UTC (rev 1394) @@ -93,10 +93,6 @@ 2 92 port-wb \ Enable A20 \ External A20GATE is disabled automatically because the internal KBC is on by default (cfg reg 8851 bit 0) - acpi-io-base 1 + 8888 config-ww \ Set ACPI base address -\ 88 8881 config-wb \ Enable ACPI regs, 32-bit PM timer, guard RTC against power glitches - 8c 8881 config-wb \ Enable ACPI regs, 32-bit PM timer, guard RTC against power glitches - fload ${BP}/cpu/x86/pc/olpc/via/startcacheasram.fth \ cache_as_ram_auto.c: amd64_main() @@ -122,10 +118,14 @@ fload ${BP}/cpu/x86/pc/olpc/via/starthostctl.fth + acpi-io-base 1 + 8888 config-ww \ Set ACPI base address + [ifdef] debug-startup cominit #) call [then] + 8c 8881 config-wb \ Enable ACPI regs, 32-bit PM timer, guard RTC against power glitches + char + report long-offsets on @@ -138,8 +138,7 @@ char X report begin hlt again then - carret report linefeed report - char F report + carret report linefeed report char F report fload ${BP}/cpu/x86/pc/olpc/via/demodram.fth then @@ -182,7 +181,7 @@ 0 [if] long-offsets on -acpi-io-base 48 + port-rl h# 1000.0000 # ax and 0<> if \ Memory ID1 bit +acpi-io-base 48 + port-rl h# 1000.0000 # ax and 0<> if \ Memory ID1 bit - set for 32bit memory width mmxcr 39 3c4 port-wb ax ax xor 3c5 port-rb ax mmxdot Modified: cpu/x86/pc/olpc/via/startcominit.fth =================================================================== --- cpu/x86/pc/olpc/via/startcominit.fth 2009-10-06 03:18:14 UTC (rev 1393) +++ cpu/x86/pc/olpc/via/startcominit.fth 2009-10-06 03:26:47 UTC (rev 1394) @@ -29,9 +29,7 @@ \ First we check for a cached board ID in CMOS RAM, to avoid the \ possibly time-consuming operation of asking the EC. - \ To read the high half of CMOS RAM we must enable it d# 17 0 devfunc - 4e 18 18 mreg \ Enable ports 74/75 for CMOS RAM access - 10 res be like Phx end-table \ Configure the I/O decoding to enable access to the EC @@ -46,24 +44,27 @@ 64 0f 07 mreg \ PCS0 size is 8 bytes - to include 68 and 6c 66 01 01 mreg \ PCS0 Enable 67 10 10 mreg \ PCS0 to LPC Bus + \ When the RTC is guarded against power glitches, there is a 50 mS delay before + \ reads return the correct data upon wakeup from S3. On power-up, the delay is + \ present even with the RTC unguarded. Unguarding it here speeds up S3 startup. + 81 ff 88 mreg \ Enable ACPI regs, 32-bit PM timer, disable RTC power glitch guard + 4e 08 08 mreg \ Enable ports 74/75 for high-bank CMOS RAM access end-table - d# 17 0 devfunc + d# 17 7 devfunc 70 fb 82 mreg \ CPU to PCI flow control - CPU to PCI posted write, Enable Delay Transaction end-table -h# 83 # al mov al h# 74 # out h# 75 # al in \ check byte - should be ~board-id -ax bx mov + acpi-io-base 4 + port-rw \ Get APCI Status register + d# 10 # ax shr 7 # ax and 1 # ax cmp <> if \ Type 1 is wakeup from S3 + \ The following applies only to power-up; no delay is needed for wakeup from S3 + \ This delay is empirically necessary before reading CMOS - minimum is 36000 - about 50 ms + \ Before the delay has elapsed, the CMOS RAM returns 0 instead of the stored value. + d# 40000 wait-us + then - \ This delay is empirically necessary before reading CMOS - minimum is 36000 - about 50 ms - \ Before the delay has elapsed, the CMOS RAM returns 0 instead of the stored value. - d# 40000 wait-us - -h# 88 # al mov al h# 74 # out bx ax mov al h# 75 # out \ Test for CMOS read timing - \ As an optimization to avoid long waits for the EC to respond, read the board ID - \ that is cached in CMOS RAM. This might not in fact be an optimization in light - \ of the above delay ... + \ that is cached in CMOS RAM. h# 83 # al mov al h# 74 # out h# 75 # al in \ check byte - should be ~board-id al ah mov ah not \ ~check byte in AH h# 82 # al mov al h# 74 # out h# 75 # al in \ board-id in AL

1 0

r1393 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Oct. 5, 2009

Oct. 5, 2009

Author: wmb Date: 2009-10-06 05:18:14 +0200 (Tue, 06 Oct 2009) New Revision: 1393 Modified: cpu/x86/pc/olpc/via/demodram.fth cpu/x86/pc/olpc/via/startgfxinit.fth cpu/x86/pc/olpc/via/starthostctl.fth Log: Via 32-bit memory width - final register value fixups as stipulated by Via. Modified: cpu/x86/pc/olpc/via/demodram.fth =================================================================== --- cpu/x86/pc/olpc/via/demodram.fth 2009-10-02 22:10:09 UTC (rev 1392) +++ cpu/x86/pc/olpc/via/demodram.fth 2009-10-06 03:18:14 UTC (rev 1393) @@ -389,17 +389,15 @@ 4b ff rank-base3 mreg \ Rank base 3 end-table - acpi-io-base 48 + port-rl h# 1000.0000 # ax and 0<> if \ Memory ID1 bit + acpi-io-base 48 + port-rl h# 1000.0000 # ax and 0<> if \ Memory ID1 bit - set for 32bit memory width 0 3 devfunc 6c 20 20 mreg \ Enable 32-bit memory width mode - channel A d4 30 10 mreg \ ODT off for low 32 bits end-table - - 0388 config-rb ax bx mov 1 # bx shr 0385 config-setup bx ax mov al dx out \ Copy Low Top from RO reg 88 to reg 85 - else - 0388 config-rb ax bx mov 0385 config-setup bx ax mov al dx out \ Copy Low Top from RO reg 88 to reg 85 then + 0388 config-rb ax bx mov 0385 config-setup bx ax mov al dx out \ Copy Low Top from RO reg 88 to reg 85 + h# 15 port80 0 [if] \ Very simple memtest Modified: cpu/x86/pc/olpc/via/startgfxinit.fth =================================================================== --- cpu/x86/pc/olpc/via/startgfxinit.fth 2009-10-02 22:10:09 UTC (rev 1392) +++ cpu/x86/pc/olpc/via/startgfxinit.fth 2009-10-06 03:18:14 UTC (rev 1393) @@ -59,7 +59,12 @@ 6f 3c4 port-wb 00 3c5 port-wb \ Base address [47:37] of SL in System Memory [then] [ifdef] xo-board - 385 config-rb \ AX: totalsize/16M + 385 config-rb ax bx mov \ BX: totalsize/16M + acpi-io-base 48 + port-rl h# 1000.0000 # ax and 0<> if \ Memory ID1 bit - set for 32bit memory width + 1 # bx shr \ Divide size by 2 for 32-bit DRAM width + then + bx ax mov \ AX: totalsize/16M_adjusted + d# 24 # ax shl \ AX: totalsize /fbmem # ax sub \ AX: totalsize-fbmemsize = fbbase d# 21 # ax shr \ AX: fbbase/2M Modified: cpu/x86/pc/olpc/via/starthostctl.fth =================================================================== --- cpu/x86/pc/olpc/via/starthostctl.fth 2009-10-02 22:10:09 UTC (rev 1392) +++ cpu/x86/pc/olpc/via/starthostctl.fth 2009-10-06 03:18:14 UTC (rev 1393) @@ -67,7 +67,7 @@ [then] end-table - acpi-io-base 48 + port-rl h# 1000.0000 # ax and 0<> if \ Memory ID1 bit + acpi-io-base 48 + port-rl h# 1000.0000 # ax and 0<> if \ Memory ID1 bit - set for 32bit memory width 0 2 devfunc 55 02 02 mreg \ Host controller to DRAM read cycle control II - 1 means 2T slower

1 0