September 2009 - openfirmware - mail.coreboot.org

r1351 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Sept. 5, 2009

Sept. 5, 2009

Author: wmb Date: 2009-09-05 11:09:26 +0200 (Sat, 05 Sep 2009) New Revision: 1351 Modified: cpu/x86/pc/olpc/via/startcominit.fth Log: Via - set the state of the WLAN activity LED in early startup. Modified: cpu/x86/pc/olpc/via/startcominit.fth =================================================================== --- cpu/x86/pc/olpc/via/startcominit.fth 2009-09-05 09:07:41 UTC (rev 1350) +++ cpu/x86/pc/olpc/via/startcominit.fth 2009-09-05 09:09:26 UTC (rev 1351) @@ -67,13 +67,18 @@ \ Now AL contains the board ID h# d1 # al cmp u< if + acpi-io-base h# 4c + port-rl h# 200000 bitclr ax dx out \ Turn off WLAN activity LED (GPIO10) + \ A-test d# 17 0 devfunc 9b 01 01 mreg \ 1 selects GPO11/12 instead of CR_PWSEL/CR_PWOFF (DCONLOAD) 46 c0 40 mreg \ Enable UART on VCP port end-table + else \ B-test or later + acpi-io-base h# 4c + port-rl h# 20000 bitset ax dx out \ Turn off WLAN activity LED (GPIO10) + \ For B-test and later, we only enable serial if the jumper is present acpi-io-base 48 + port-rb h# 10 # al test 0= if d# 17 0 devfunc

1 0

r1350 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Sept. 5, 2009

Sept. 5, 2009

Author: wmb Date: 2009-09-05 11:07:41 +0200 (Sat, 05 Sep 2009) New Revision: 1350 Modified: cpu/x86/pc/olpc/via/romreset.bth cpu/x86/pc/olpc/via/startcacheasram.fth Log: Via - speed up startup by eliminating unnecessary cache prefill from slow ROM, but add back just the right amount of explicit delay to ensure that the system starts properly. Modified: cpu/x86/pc/olpc/via/romreset.bth =================================================================== --- cpu/x86/pc/olpc/via/romreset.bth 2009-09-05 09:05:43 UTC (rev 1349) +++ cpu/x86/pc/olpc/via/romreset.bth 2009-09-05 09:07:41 UTC (rev 1350) @@ -92,7 +92,7 @@ fload ${BP}/cpu/x86/pc/olpc/via/startcacheasram.fth - \ We are now running from cache and can use the stack + d# 40000 wait-us \ This delay is empirically necessary - minimum is 36000 - about 50 ms \ cache_as_ram_auto.c: amd64_main() @@ -131,7 +131,7 @@ \ This turns an OS reboot into a real cold start - from coreboot h# 380 config-rb ax ax or 0<> if \ C-page shadowing not on yet 6 h# cf9 port-wb \ Force a full system reset - char X report + char X report begin hlt again then char F report @@ -169,6 +169,7 @@ fload ${BP}/cpu/x86/pc/olpc/via/startmtrrinit.fth fload ${BP}/cpu/x86/pc/olpc/via/ioinit.fth +\ XXX need to move startcpuspeed to here acpi-io-base 4 + port-rw \ Get APCI Status register ax bx mov Modified: cpu/x86/pc/olpc/via/startcacheasram.fth =================================================================== --- cpu/x86/pc/olpc/via/startcacheasram.fth 2009-09-05 09:05:43 UTC (rev 1349) +++ cpu/x86/pc/olpc/via/startcacheasram.fth 2009-09-05 09:07:41 UTC (rev 1350) @@ -26,16 +26,6 @@ cld - \ Access ROM to load it into the icache - dropin-base # esi mov - /icached 4 / # ecx mov - rep eax lods - - \ Access "RAM" area to load it into the dcache - dcached-base # esi mov - /dcached 4 / # ecx mov - rep eax lods - \ Put the stack pointer at the top of the dcached area dcached-base /dcached + 4 - # esp mov ds ax mov ax ss mov

1 0

r1349 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Sept. 5, 2009

Sept. 5, 2009

Author: wmb Date: 2009-09-05 11:05:43 +0200 (Sat, 05 Sep 2009) New Revision: 1349 Modified: cpu/x86/pc/olpc/via/ioinit.fth Log: OLPC trac 9449 - leave PLL on in suspend state, thus preventing random reboots on the first resume. Also set the mux for the SD power control GPIO correctly. Modified: cpu/x86/pc/olpc/via/ioinit.fth =================================================================== --- cpu/x86/pc/olpc/via/ioinit.fth 2009-09-05 09:03:28 UTC (rev 1348) +++ cpu/x86/pc/olpc/via/ioinit.fth 2009-09-05 09:05:43 UTC (rev 1349) @@ -43,6 +43,7 @@ 89 ff f8 mreg \ Dynamic clocks 8b ff bf mreg \ Dynamic clocks 8d ff 30 mreg \ Self-refresh in C3 and C4 + 8e ff 20 mreg \ Leave PLL on in suspend state - necessary for reliable S3 90 ff ff mreg \ Gate clocks 91 ff ff mreg \ Gate clocks 92 cc cc mreg \ Dynamic buffer control, power down comparators @@ -222,7 +223,7 @@ [then] [ifdef] xo-board - 9b ff 88 mreg \ 80 undoc but is LVDS power. 00 forces LVDS power off, 80 lets 3d5.D2[7,6,3] control it; 1 selects GPO11/12 instead of CR_PWSEL/CR_PWOFF (DCONLOAD) + 9b fe 88 mreg \ 80 undoc but is LVDS power. 00 forces LVDS power off, 80 lets 3d5.D2[7,6,3] control it; 1 selects GPO11/12 instead of CR_PWSEL/CR_PWOFF (DCONLOAD) 9f ff 08 mreg \ Slot 3 is SDIO, no pullup on KB/MS, fastest SD [then] [ifdef] demo-board

1 0

r1348 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Sept. 5, 2009

Sept. 5, 2009

Author: wmb Date: 2009-09-05 11:03:28 +0200 (Sat, 05 Sep 2009) New Revision: 1348 Modified: cpu/x86/pc/olpc/via/boardrev.fth Log: Via - cache the board ID value in CMOS as the raw EC code, not the derived ID, thus making the assembly language startup code easier. Modified: cpu/x86/pc/olpc/via/boardrev.fth =================================================================== --- cpu/x86/pc/olpc/via/boardrev.fth 2009-09-05 09:02:27 UTC (rev 1347) +++ cpu/x86/pc/olpc/via/boardrev.fth 2009-09-05 09:03:28 UTC (rev 1348) @@ -27,23 +27,13 @@ : stand-init-io stand-init-io -[ifndef] notdef - ['] board-id@ catch if 0 then case - 0 of 0 endof \ EC broken - ( board-id ) dup h# 10 * 8 + swap \ E.g. b3 -> b38 - endcase -[else] - \ wait-ib-empty - \ autowack-off - 0 h# 68 pc! - h# d08 -[then] + ['] board-id@ catch if 0 then - to board-revision - \ Cache the board revision in CMOS RAM so the early startup code \ can get it without having to wait for the EC. - board-revision dup h# 82 cmos! invert h# 83 cmos! + dup dup h# 82 cmos! invert h# 83 cmos! ( id ) + + dup if h# 10 * 8 + then to board-revision ; warning !

1 0

r1347 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Sept. 5, 2009

Sept. 5, 2009

Author: wmb Date: 2009-09-05 11:02:27 +0200 (Sat, 05 Sep 2009) New Revision: 1347 Modified: cpu/x86/pc/olpc/via/acpi.fth Log: Via - when clearing power management events, write the register as a word not a long, thus avoiding "leakage" into the enable register. Modified: cpu/x86/pc/olpc/via/acpi.fth =================================================================== --- cpu/x86/pc/olpc/via/acpi.fth 2009-09-03 05:18:39 UTC (rev 1346) +++ cpu/x86/pc/olpc/via/acpi.fth 2009-09-05 09:02:27 UTC (rev 1347) @@ -299,7 +299,7 @@ [then] 4 acpi-w@ 1 or 4 acpi-w! \ Set SCI_EN bit - h# ffffffff h# 20 acpi-l! \ Ack all leftover events + h# ffffffff h# 20 acpi-w! \ Ack all leftover events ; : acpi-power-off ( -- )

1 0

r1346 - cpu/x86/pc/olpc
by svn＠openfirmware.info Sept. 2, 2009

Sept. 2, 2009

Author: wmb Date: 2009-09-03 07:18:39 +0200 (Thu, 03 Sep 2009) New Revision: 1346 Modified: cpu/x86/pc/olpc/versions.fth Log: OLPC XO-1 versions.fth - Use new version of nandblaster support routines, related to trac tickets 9434 and 9045. Modified: cpu/x86/pc/olpc/versions.fth =================================================================== --- cpu/x86/pc/olpc/versions.fth 2009-09-03 05:16:25 UTC (rev 1345) +++ cpu/x86/pc/olpc/versions.fth 2009-09-03 05:18:39 UTC (rev 1346) @@ -2,7 +2,7 @@ \ The overall firmware revision macro: FW_MAJOR E -macro: FW_MINOR 41c +macro: FW_MINOR 41d \ The EC microcode macro: EC_VERSION e34 @@ -11,8 +11,8 @@ \ Temporarily uncomment the line and modify the path as necessary \ macro: GET_EC cp pq2e18c.img ec.img -macro: KEYS mpkeys -\ macro: KEYS testkeys +\ macro: KEYS mpkeys +macro: KEYS testkeys \ The wireless LAN module firmware macro: WLAN_RPM ${WLAN_VERSION}.olpc1 @@ -26,6 +26,6 @@ \ With a specific ID, mcastnand.bth will download a tarball without .git stuff. \ With "test", mcastnand.bth will clone the git head if build/multicast-nand/ \ is not already present, then you can modify the git subtree as needed. -macro: MCNAND_VERSION b9a9d22b6037c3891f9cf8eabeaf7cd9efbd5241 +macro: MCNAND_VERSION 74bf48045bca9d4919728a7bbd7a0acb96c0d8ef \ macro: MCNAND_VERSION test \ macro: MCNAND_VERSION HEAD

1 0

r1345 - cpu/x86/pc/olpc/via
by svn＠openfirmware.info Sept. 2, 2009

Sept. 2, 2009

Author: wmb Date: 2009-09-03 07:16:25 +0200 (Thu, 03 Sep 2009) New Revision: 1345 Modified: cpu/x86/pc/olpc/via/ioinit.fth cpu/x86/pc/olpc/via/startcominit.fth Log: Via - yet another early startup change to handle A-test vs B-test differences, this time having to do with the SDIO power switch for SDIO slot 2. Modified: cpu/x86/pc/olpc/via/ioinit.fth =================================================================== --- cpu/x86/pc/olpc/via/ioinit.fth 2009-09-03 05:14:31 UTC (rev 1344) +++ cpu/x86/pc/olpc/via/ioinit.fth 2009-09-03 05:16:25 UTC (rev 1345) @@ -222,7 +222,7 @@ [then] [ifdef] xo-board - 9b ff 89 mreg \ 80 undoc but is LVDS power. 00 forces LVDS power off, 80 lets 3d5.D2[7,6,3] control it; 1 selects GPO11/12 instead of CR_PWSEL/CR_PWOFF (DCONLOAD) + 9b ff 88 mreg \ 80 undoc but is LVDS power. 00 forces LVDS power off, 80 lets 3d5.D2[7,6,3] control it; 1 selects GPO11/12 instead of CR_PWSEL/CR_PWOFF (DCONLOAD) 9f ff 08 mreg \ Slot 3 is SDIO, no pullup on KB/MS, fastest SD [then] [ifdef] demo-board Modified: cpu/x86/pc/olpc/via/startcominit.fth =================================================================== --- cpu/x86/pc/olpc/via/startcominit.fth 2009-09-03 05:14:31 UTC (rev 1344) +++ cpu/x86/pc/olpc/via/startcominit.fth 2009-09-03 05:16:25 UTC (rev 1345) @@ -24,64 +24,64 @@ \ If the SERIAL_EN jumper is installed, or if the machine is an A-test, \ route the external pin to the UART; otherwise leave it connected to the VCP port. - acpi-io-base 48 + port-rb h# 10 # al test 0= if - \ SERIAL_EN is installed, so put the A-test ID in AL so we can check - \ a single condition at the end of the logic. - h# d0 # al mov - else - \ SERIAL_EN is not installed. Determine the board ID. + \ SERIAL_EN is not installed. Determine the board ID. - \ First we check for a cached board ID in CMOS RAM, to avoid the - \ possibly time-consuming operation of asking the EC. + \ 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 + \ 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 - 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 + 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 - al ah cmp 0<> if \ If the check byte matches, fall through with the ID in AL + al ah cmp 0<> if \ If the check byte matches, fall through with the ID in AL - \ If check byte is wrong, we have to ask the EC + \ If check byte is wrong, we have to ask the EC - h# 6c # al in \ EC status register - 2 # al and \ input buffer full bit - 0<> if \ If the bit is nonzero, we can't send a command yet - \ We don't wait for the EC; if it is busy we assume B-test - \ It shouldn't be busy at this point because we haven't tried to talk to it yet - h# d1 # al mov \ EC busy - report B-test + h# 6c # al in \ EC status register + 2 # al and \ input buffer full bit + 0<> if \ If the bit is nonzero, we can't send a command yet + \ We don't wait for the EC; if it is busy we assume B-test + \ It shouldn't be busy at this point because we haven't tried to talk to it yet + h# d1 # al mov \ EC busy - report B-test + else + h# 19 # al mov al h# 6c # out \ Send board ID command to EC + d# 200 # cx mov \ Wait up to 200 mS for the EC to respond + begin + d# 1000 wait-us \ 1 mS delay so we don't pound on the EC + h# 6c # al in \ Get status register + 3 # al and \ Check for output buffer full + 1 # al cmp + loopne + <> if \ Not equal means timeout + h# d1 # al mov \ EC timeout - report B-test else - h# 19 # al mov al h# 6c # out \ Send board ID command to EC - d# 200 # cx mov \ Wait up to 200 mS for the EC to respond - begin - d# 1000 wait-us \ 1 mS delay so we don't pound on the EC - h# 6c # al in \ Get status register - 3 # al and \ Check for output buffer full - 1 # al cmp - loopne - <> if \ Not equal means timeout - h# d1 # al mov \ EC timeout - report B-test - else - h# 68 # al in \ Get board ID byte from EC - then + h# 68 # al in \ Get board ID byte from EC then then then - \ Now AL contains the board ID or a surrogate that will cause the - \ desired result for the test below. + \ Now AL contains the board ID h# d1 # al cmp u< if + \ A-test d# 17 0 devfunc - \ The following is for UART on VCP port - 46 c0 40 mreg - \ The following is for UART on DVP port - \ 46 c0 c0 mreg + 9b 01 01 mreg \ 1 selects GPO11/12 instead of CR_PWSEL/CR_PWOFF (DCONLOAD) + 46 c0 40 mreg \ Enable UART on VCP port end-table + else + \ B-test or later + \ For B-test and later, we only enable serial if the jumper is present + acpi-io-base 48 + port-rb h# 10 # al test 0= if + d# 17 0 devfunc + 46 c0 40 mreg \ Enable UART on VCP port + end-table + then then - + d# 17 0 devfunc \ Standard COM2 and COM1 IRQ routing b2 ff 34 mreg

1 0

r1344 - cpu/x86/pc/olpc
by svn＠openfirmware.info Sept. 2, 2009

Sept. 2, 2009

Author: wmb Date: 2009-09-03 07:14:31 +0200 (Thu, 03 Sep 2009) New Revision: 1344 Modified: cpu/x86/pc/olpc/crypto.fth Log: OLPC crypto.fth - fixed a bug that was causing false failures of the crypto-hash function as a result of some extra status bits that were added to the return code. Modified: cpu/x86/pc/olpc/crypto.fth =================================================================== --- cpu/x86/pc/olpc/crypto.fth 2009-09-03 05:13:18 UTC (rev 1343) +++ cpu/x86/pc/olpc/crypto.fth 2009-09-03 05:14:31 UTC (rev 1344) @@ -36,7 +36,7 @@ d# 128 buffer: hashbuf : crypto-hash ( data$ hashname$ -- result$ ) 2>r 0 -rot hashbuf d# 128 hashlen 0 2r> ( 0 data$ sig$ key$ hashname$ ) - signature-bad? abort" Hash failed" ( ) + signature-bad? h# fffff and abort" Hash failed" ( ) hashbuf hashlen @ ;

1 0

r1343 - cpu/x86/pc/olpc
by svn＠openfirmware.info Sept. 2, 2009

Sept. 2, 2009

Author: wmb Date: 2009-09-03 07:13:18 +0200 (Thu, 03 Sep 2009) New Revision: 1343 Modified: cpu/x86/pc/olpc/security.fth Log: OLPC security.fth - commented out some words that are no longer used now that we support key delegation. Modified: cpu/x86/pc/olpc/security.fth =================================================================== --- cpu/x86/pc/olpc/security.fth 2009-09-03 05:11:26 UTC (rev 1342) +++ cpu/x86/pc/olpc/security.fth 2009-09-03 05:13:18 UTC (rev 1343) @@ -390,22 +390,6 @@ : my-sn$ ( -- adr len ) machine-id-buf d# 11 ; - -\ parse-sig parses a "sig01:" format signature string, returning its -\ hashname and signature substrings. It converts the signature -\ substring from ASCII hex to binary bytes. - -: parse-sig ( sig01$ -- true | hashname$ sig$ false ) - dup d# 89 < if 2drop true exit then - bl left-parse-string " sig01:" $= 0= if 2drop true exit then ( rem$ ) - bl left-parse-string dup d# 6 <> if 4drop true exit then 2swap ( hash$ rem$ ) - bl left-parse-string nip d# 64 <> if 4drop true exit then ( hash$ rem$ ) - newline left-parse-string 2swap nip 0<> if 4drop true exit then ( hash$ data$ ) - dup /sig 2* <> if ( ." Bad signature length" cr ) 2drop true exit then ( hash$ data$ ) - - hex-decode if 2drop true else false then -; - \ zip-extent looks inside a memory-resident ZIP archive and returns \ the address,length of a given component of that archive. This \ assumes that the components are "stored", not "deflated". It @@ -446,10 +430,22 @@ true ; -\ hashname remembers the most recently used hashname to guard against -\ attacks based on reuse of the same (presumably compromized) hash. - 0 [if] +\ parse-sig parses a "sig01:" format signature string, returning its +\ hashname and signature substrings. It converts the signature +\ substring from ASCII hex to binary bytes. + +: parse-sig ( sig01$ -- true | hashname$ sig$ false ) + dup d# 89 < if 2drop true exit then + bl left-parse-string " sig01:" $= 0= if 2drop true exit then ( rem$ ) + bl left-parse-string dup d# 6 <> if 4drop true exit then 2swap ( hash$ rem$ ) + bl left-parse-string nip d# 64 <> if 4drop true exit then ( hash$ rem$ ) + newline left-parse-string 2swap nip 0<> if 4drop true exit then ( hash$ data$ ) + dup /sig 2* <> if ( ." Bad signature length" cr ) 2drop true exit then ( hash$ data$ ) + + hex-decode if 2drop true else false then +; + \ signature-invalid? checks the validity of data$ against the ASCII signature \ record sig0N$, using the public key that thiskey$ points to. \ It also verifies that the hashname contained in sig01$ is the @@ -478,6 +474,9 @@ ; [then] +\ exp-hashname$ remembers the most recently used hashname to guard against +\ attacks based on reuse of the same (presumably compromized) hash. + 0 0 2value exp-hashname$ 0 0 2value signed-data$ @@ -621,9 +620,8 @@ ; \ Find a sig0N: line and check its sha256/rsa signature -: sha-valid? ( data$ sig01$ -- okay? ) - next-sig-in-list$ if 2drop false exit then ( data$ rem$ sig01$ ) - \ 2nip " sha256" signature-invalid? 0= +: sha-valid? ( data$ sig$ -- okay? ) + next-sig-in-list$ if 2drop false exit then ( data$ rem$ sig$ ) 2nip " sha256" signature-good? ; @@ -631,12 +629,11 @@ \ and check their signatures : fw-valid? ( data$ sig$ -- okay? ) 2swap 2>r ( sig$ r: data$ ) - next-sig-in-list$ if 2r> 2drop false exit then ( rem$ sig01$ ) + next-sig-in-list$ if 2r> 2drop false exit then ( rem$ sig$ ) 2r@ 2swap sha-valid? 0= if ( rem$ r: data$ ) 2r> 4drop false exit then ( rmd-sig$ r: data$ ) - next-sig$ if 2r> 2drop false exit then ( rem$ sig01$ ) -\ 2nip 2r> 2swap " rmd160" signature-invalid? 0= + next-sig$ if 2r> 2drop false exit then ( rem$ sig$ ) 2nip 2r> 2swap " rmd160" signature-good? ;

1 0

r1342 - dev
by svn＠openfirmware.info Sept. 2, 2009

Sept. 2, 2009

Author: wmb Date: 2009-09-03 07:11:26 +0200 (Thu, 03 Sep 2009) New Revision: 1342 Modified: dev/libertas.fth Log: Libertas driver - eliminate or at least greatly reduce "marvel-get-mac-address failed" messages by retrying that command, and by not calling it after the mac address is already known. Also deassociate on a force open, so find-multinand-server works correctly after the driver has already been opened once in normal association mode. Modified: dev/libertas.fth =================================================================== --- dev/libertas.fth 2009-09-02 23:40:15 UTC (rev 1341) +++ dev/libertas.fth 2009-09-03 05:11:26 UTC (rev 1342) @@ -580,12 +580,21 @@ \ MAC address \ ========================================================================= -: marvel-get-mac-address ( -- ) +\ This command has an annoying tendency to fail to sometimes - wait-cmd-resp +\ times out. Retrying usually fixes it. +: (marvel-get-mac-address) ( -- error? ) 8 h# 4d ( CMD_802_11_MAC_ADDRESS ) prepare-cmd ACTION_GET +xw - outbuf-wait if ." marvel-get-mac-address failed" cr exit then + outbuf-wait if true exit then respbuf >fw-data 2 + mac-adr$ move + false ; +: marvel-get-mac-address ( -- ) + 4 0 do + (marvel-get-mac-address) 0= if unloop exit then + loop + ." marvel-get-mac-address failed" cr +; : marvel-set-mac-address ( -- ) 8 h# 4d ( CMD_802_11_MAC_ADDRESS ) prepare-cmd @@ -1516,6 +1525,7 @@ : ?make-mac-address-property ( -- ) driver-state ds-ready < if exit then " mac-address" get-my-property if + marvel-get-mac-address mac-adr$ encode-bytes " local-mac-address" property mac-address encode-bytes " mac-address" property else @@ -1531,7 +1541,6 @@ ; : init-net ( -- ) - marvel-get-mac-address ?make-mac-address-property ; @@ -1576,7 +1585,9 @@ my-args " supplicant" $open-package to supplicant-ih supplicant-ih 0= if release-bus-resources free-buf false exit then nonce-cmd - force-open? 0= if + force-open? if + ds-disconnected reset-driver-state + else link-up? 0= if ['] 2drop to ?process-eapol do-associate 0= if free-buf false exit then

1 0