OpenBIOS January 2014

openbios@openbios.org

7 participants
15 discussions

Some SPARC32 Patches
by Olivier Danet 31 Jan '14

31 Jan '14

Hello ! Here are a few patches for QEMU and OpenBIOS, about SPARC32, grouped into two diffs: OpenBIOS : http://temlib.org/pub/openbios_nn.diff QEMU : http://temlib.org/pub/qemu_nn.diff Changes: OpenBIOS * SparcStation10/20 mode fix (drivers/iommu.c, drivers/sbus.c) - TCX properties (drivers/tcx.fs) - next-property (forth/device/property.fs) - Keyboard property (drivers/escc.c) + Memory allocation/deallocation (arch/sparc32/lib.c, arch/sparc32/ofmem_sparc32.c, libopenbios/ofmem_common.c, arch/sparc32/romvec.h) + a.out file loading (libopenbios/aout_load.c) + Dummy TICK counter (arch/sparc32/romvec.c) + Interrupt/Timer properties (include/drivers/drivers.h, forth/device/property.fs, drivers/obio.c) QEMU - TCX Hardware acceleration : Blitter, Stippler, HW cursor (hw/display/tcx.c, hw/sparc/sun4m.c) - "!TC on DATA XFER" warning under NetBSD (hw/scsi/esp.c) + Timer Run/Stop bit correction (hw/timer/slavio_timer.c) TCX emulation with acceleration seems good enough, although it is incomplete and it is probably impossible to implement the 24bits mode accurately and efficiently in QEMU. I will post each patch separately on the two mailing lists. TCX patches should be synchronised : OpenBIOS cannot advertise an hardware cursor until it is actually implemented. One more thing. After NetBSD and TCX acceleration, I tried NextSTEP, which needs the patches marked + above. Alas, this is not enough to make it run. Maybe it is because QEMU SCSI disk and ESP emulation are too fast. Olivier

5 4

Re: [OpenBIOS] next-property
by Olivier Danet 25 Jan '14

25 Jan '14

On 03/01/2014 17:09, Artyom Tarasenko wrote: > On Thu, Jan 2, 2014 at 2:43 AM, Olivier Danet <odanet(a)caramail.com> wrote: >> I may have found a bug in OpenBIOS. >> >> An OS uses the "nextprop" iterator to scan device properties, and eventually >> asks about nonexistant properties. >> "const char *obp_nextprop(int node, const char *name)" >> (arch/sparc32/romvec.c) >> calls the "next-property" forth code, then pops the result. >> According to the standard, next-property should return 0 if the property >> selected >> is the last _OR_ if the property does not exist. >> The current code in forth/admin/devices.fs returns nothing when the property >> does >> not exit. >> >> For example (SPARC 32) : >> >> cd screen >> " name" ?active-package next-property drop cr type >> --> returns "device_type" >> >> " interrupts" ?active-package next-property >> --> returns 0, this is the last property. >> >> " bozo" ?active-package next-property >> --> returns nothing. Which, imho, is wrong. > Actually this matches pretty much with what OBP does: > > ok showstack > ok " model" current-device next-property > fffe61e8 4 ffffffffffffffff ok drop type > name > ok " bozo" current-device next-property > ok > > Artyom > >> Here is a patch, I am not quite sure about it (debuting Forth) >> but it helps the target OS : >> >> Index: forth/device/property.fs >> =================================================================== >> --- forth/device/property.fs (révision 1246) >> +++ forth/device/property.fs (copie de travail) >> @@ -70,7 +70,7 @@ >> 2drop r> >dn.properties @ >> else >> r> find-property dup if @ then >> - ?dup if >prop.next @ then >> + dup if >prop.next @ then >> then >> >> ?dup if >> =================================================================== >> >> The standard asks for a pointer to a zero lenght string instead of zero, >> the C code obp_nextprop() expects a zero. I don't know if it makes a >> difference. >> >> Olivier >> (Btw, the OS is NetBSD :-) >> >> >> -- >> OpenBIOS http://openbios.org/ >> Mailinglist: http://lists.openbios.org/mailman/listinfo >> Free your System - May the Forth be with you > > You are right. I tried on an Ultra 10 with OpenBoot 3.12. next-property returns nothing when the property is unknown and also when the property is the last one. This contradicts Sun own documentation "Writing FCode 3.x Programs". Puzzled, I tried then on a Apple iBook G3 : dev screen " <something>" active-package next-property This laptop returns : - true and a forth string when the property is not the last one - true and "0 0" when the property is the last one - false when the property does not exist. Anyway, whatever method is chosen (return nothing, return zero...), OpenBIOS forth and/or C code ("opb_nextprop()") should be modifed to avoid page faults. Olivier

3 10

[commit] r1257 - trunk/openbios-devel/arch/sparc32
by repository service 13 Jan '14

13 Jan '14

Author: mcayland Date: Mon Jan 13 10:47:06 2014 New Revision: 1257 URL: http://tracker.coreboot.org/trac/openbios/changeset/1257 Log: SPARC32: create proper romvec memory lists from OFMEM memory properties Previously we would try and set safe values for the romvec properties list rather than calculating them. This should provide older clients with a more accurate representation of the memory areas they are allowed to use. Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> Modified: trunk/openbios-devel/arch/sparc32/boot.c trunk/openbios-devel/arch/sparc32/lib.c trunk/openbios-devel/arch/sparc32/romvec.c Modified: trunk/openbios-devel/arch/sparc32/boot.c ============================================================================== --- trunk/openbios-devel/arch/sparc32/boot.c Mon Jan 13 10:47:04 2014 (r1256) +++ trunk/openbios-devel/arch/sparc32/boot.c Mon Jan 13 10:47:06 2014 (r1257) @@ -8,6 +8,7 @@ #include "drivers/drivers.h" #include "libc/diskio.h" #include "libc/vsprintf.h" +#include "libopenbios/ofmem.h" #include "libopenbios/sys_info.h" #include "openprom.h" #include "boot.h" @@ -20,17 +21,20 @@ char boot_device; const void *romvec; +static struct linux_mlist_v0 *totphyslist, *availlist, *prommaplist; static void setup_romvec(void) { /* SPARC32 is slightly unusual in that before invoking any loaders, a romvec array needs to be set up to pass certain parameters using a C struct. Hence this function extracts the relevant boot information and places it in obp_arg. */ - - int intprop, proplen, target, device; + + int intprop, proplen, target, device, i; + unsigned int *intprop_ptr; phandle_t chosen; char *prop, *id, *name; static char bootpathbuf[128], bootargsbuf[128], buf[128]; + struct linux_mlist_v0 **pp; /* Get the stdin and stdout paths */ chosen = find_dev("/chosen"); @@ -132,6 +136,60 @@ obp_arg.argv[1] = bootargsbuf; } + + /* Generate the totphys (total memory available) list */ + prop = get_property(s_phandle_memory, "reg", &proplen); + intprop_ptr = (unsigned int *)prop; + + for (pp = &totphyslist, i = 0; i < (proplen / sizeof(int)); pp = &(**pp).theres_more, i+=3) { + *pp = (struct linux_mlist_v0 *)malloc(sizeof(struct linux_mlist_v0)); + (**pp).theres_more = NULL; + (**pp).start_adr = (char *)intprop_ptr[1]; + (**pp).num_bytes = intprop_ptr[2]; + + printk("start_adr: %x\n", (int)(**pp).start_adr); + printk("num_bytes: %x\n", (**pp).num_bytes); + + intprop_ptr += 3; + } + + /* Generate the avail (physical memory available) list */ + prop = get_property(s_phandle_memory, "available", &proplen); + intprop_ptr = (unsigned int *)prop; + + for (pp = &availlist, i = 0; i < (proplen / sizeof(int)); pp = &(**pp).theres_more, i+=3) { + *pp = (struct linux_mlist_v0 *)malloc(sizeof(struct linux_mlist_v0)); + (**pp).theres_more = NULL; + (**pp).start_adr = (char *)intprop_ptr[1]; + (**pp).num_bytes = intprop_ptr[2]; + + intprop_ptr += 3; + } + + /* Generate the prommap (taken virtual memory) list from inverse of available */ + prop = get_property(s_phandle_mmu, "available", &proplen); + intprop_ptr = (unsigned int *)prop; + + for (pp = &prommaplist, i = 0; i < (proplen / sizeof(int)); pp = &(**pp).theres_more, i+=3) { + *pp = (struct linux_mlist_v0 *)malloc(sizeof(struct linux_mlist_v0)); + (**pp).theres_more = NULL; + (**pp).start_adr = (char *)(intprop_ptr[1] + intprop_ptr[2]); + + if (i + 3 < (proplen / sizeof(int))) { + /* Size from next entry */ + (**pp).num_bytes = (intprop_ptr[4] + intprop_ptr[5]) - (intprop_ptr[1] + intprop_ptr[2]); + } else { + /* Tail (size from top of virtual memory) */ + (**pp).num_bytes = 0xffffffffUL - (intprop_ptr[1] + intprop_ptr[2]); + } + + intprop_ptr += 3; + } + + /* Finally set the memory properties */ + ((struct linux_romvec *)romvec)->pv_v0mem.v0_totphys = &totphyslist; + ((struct linux_romvec *)romvec)->pv_v0mem.v0_available = &availlist; + ((struct linux_romvec *)romvec)->pv_v0mem.v0_prommap = &prommaplist; } Modified: trunk/openbios-devel/arch/sparc32/lib.c ============================================================================== --- trunk/openbios-devel/arch/sparc32/lib.c Mon Jan 13 10:47:04 2014 (r1256) +++ trunk/openbios-devel/arch/sparc32/lib.c Mon Jan 13 10:47:06 2014 (r1257) @@ -59,14 +59,6 @@ unsigned long *l1; static unsigned long *context_table; -static ucell *mem_reg = 0; -static ucell *mem_avail = 0; -static ucell *virt_avail = 0; - -static struct linux_mlist_v0 totphys[1]; -static struct linux_mlist_v0 totmap[1]; -static struct linux_mlist_v0 totavail[1]; - struct linux_mlist_v0 *ptphys; struct linux_mlist_v0 *ptmap; struct linux_mlist_v0 *ptavail; @@ -200,55 +192,6 @@ ofmem_arch_map_pages(pa, va, size, ofmem_arch_default_translation_mode(pa)); } -static void -update_memory_properties(void) -{ - /* Update the device tree memory properties from the master - totphys, totmap and totavail romvec arrays */ - mem_reg[0] = 0; - mem_reg[1] = pointer2cell(totphys[0].start_adr); - mem_reg[2] = totphys[0].num_bytes; - - virt_avail[0] = 0; - virt_avail[1] = 0; - virt_avail[2] = pointer2cell(totmap[0].start_adr); - - mem_avail[0] = 0; - mem_avail[1] = pointer2cell(totavail[0].start_adr); - mem_avail[2] = totavail[0].num_bytes; -} - -static void -init_romvec_mem(void) -{ - ptphys = totphys; - ptmap = totmap; - ptavail = totavail; - - /* - * Form memory descriptors. - */ - totphys[0].theres_more = NULL; - totphys[0].start_adr = (char *) 0; - totphys[0].num_bytes = qemu_mem_size; - - totavail[0].theres_more = NULL; - totavail[0].start_adr = (char *) 0; - totavail[0].num_bytes = va2pa((int)&_start) - PAGE_SIZE; - - totmap[0].theres_more = NULL; - totmap[0].start_adr = &_start; - totmap[0].num_bytes = (unsigned long) &_iomem - - (unsigned long) &_start + PAGE_SIZE; - - /* Pointers to device tree memory properties */ - mem_reg = malloc(sizeof(ucell) * 3); - mem_avail = malloc(sizeof(ucell) * 3); - virt_avail = malloc(sizeof(ucell) * 3); - - update_memory_properties(); -} - char *obp_dumb_mmap(char *va, int which_io, unsigned int pa, unsigned int size) { @@ -326,8 +269,6 @@ phys_addr_t virtregsize; ofmem_t *ofmem = ofmem_arch_get_private(); - init_romvec_mem(); - /* Find the phandles for the /memory and /virtual-memory nodes */ push_str("/memory"); fword("find-package"); Modified: trunk/openbios-devel/arch/sparc32/romvec.c ============================================================================== --- trunk/openbios-devel/arch/sparc32/romvec.c Mon Jan 13 10:47:04 2014 (r1256) +++ trunk/openbios-devel/arch/sparc32/romvec.c Mon Jan 13 10:47:06 2014 (r1257) @@ -461,9 +461,9 @@ romvec0.pv_romvers = 3; romvec0.pv_plugin_revision = 2; romvec0.pv_printrev = 0x20019; - romvec0.pv_v0mem.v0_totphys = &ptphys; - romvec0.pv_v0mem.v0_prommap = &ptmap; - romvec0.pv_v0mem.v0_available = &ptavail; + romvec0.pv_v0mem.v0_totphys = NULL; + romvec0.pv_v0mem.v0_prommap = NULL; + romvec0.pv_v0mem.v0_available = NULL; romvec0.pv_nodeops = &nodeops0; romvec0.pv_bootstr = (void *)doublewalk; romvec0.pv_v0devops.v0_devopen = &obp_devopen_handler;

1 0

[commit] r1256 - trunk/openbios-devel/arch/sparc32
by repository service 13 Jan '14

13 Jan '14

Author: mcayland Date: Mon Jan 13 10:47:04 2014 New Revision: 1256 URL: http://tracker.coreboot.org/trac/openbios/changeset/1256 Log: SPARC32: move romvec initialisation out of go() and into a separate function The go word should just execute the current program context, otherwise the romvec contents will get altered if called by a client. Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> Modified: trunk/openbios-devel/arch/sparc32/boot.c Modified: trunk/openbios-devel/arch/sparc32/boot.c ============================================================================== --- trunk/openbios-devel/arch/sparc32/boot.c Mon Jan 13 10:46:59 2014 (r1255) +++ trunk/openbios-devel/arch/sparc32/boot.c Mon Jan 13 10:47:04 2014 (r1256) @@ -20,26 +20,18 @@ char boot_device; const void *romvec; -void go(void) + +static void setup_romvec(void) { - ucell address, type, size; - int image_retval = 0, intprop, proplen, target, device; + /* SPARC32 is slightly unusual in that before invoking any loaders, a romvec array + needs to be set up to pass certain parameters using a C struct. Hence this function + extracts the relevant boot information and places it in obp_arg. */ + + int intprop, proplen, target, device; phandle_t chosen; char *prop, *id, *name; static char bootpathbuf[128], bootargsbuf[128], buf[128]; - /* Get the entry point and the type (see forth/debugging/client.fs) */ - feval("saved-program-state >sps.entry @"); - address = POP(); - feval("saved-program-state >sps.file-type @"); - type = POP(); - feval("saved-program-state >sps.file-size @"); - size = POP(); - - /* SPARC32 is slightly unusual in that before invoking any loaders, a romvec array - needs to be set up to pass certain parameters using a C struct. Hence this section - extracts the relevant boot information and places it in obp_arg. */ - /* Get the stdin and stdout paths */ chosen = find_dev("/chosen"); intprop = get_int_property(chosen, "stdin", &proplen); @@ -140,7 +132,24 @@ obp_arg.argv[1] = bootargsbuf; } - +} + + +void go(void) +{ + ucell address, type, size; + int image_retval = 0; + + /* Get the entry point and the type (see forth/debugging/client.fs) */ + feval("saved-program-state >sps.entry @"); + address = POP(); + feval("saved-program-state >sps.file-type @"); + type = POP(); + feval("saved-program-state >sps.file-size @"); + size = POP(); + + setup_romvec(); + printk("\nJumping to entry point " FMT_ucellx " for type " FMT_ucellx "...\n", address, type); switch (type) {

1 0

[commit] r1255 - trunk/openbios-devel/libopenbios
by repository service 13 Jan '14

13 Jan '14

Author: mcayland Date: Mon Jan 13 10:46:59 2014 New Revision: 1255 URL: http://tracker.coreboot.org/trac/openbios/changeset/1255 Log: OFMEM: allocate retained memory top downwards Not only does this match the behaviour of the other routines, but it also fixes a bug whereby the first free physical memory location was 0x0 and so page zero was being returned for use by SUNW,retain clients. Due to the way in which PCI/ISA devices are mapped in QEMU (e.g. IDE ioports) then clients could inadvertantly start writing to device registers. Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> Modified: trunk/openbios-devel/libopenbios/ofmem_common.c Modified: trunk/openbios-devel/libopenbios/ofmem_common.c ============================================================================== --- trunk/openbios-devel/libopenbios/ofmem_common.c Mon Jan 13 10:46:56 2014 (r1254) +++ trunk/openbios-devel/libopenbios/ofmem_common.c Mon Jan 13 10:46:59 2014 (r1255) @@ -599,7 +599,7 @@ " align=" FMT_ucellx "\n", phys, size, align); - retain_phys = ofmem_claim_phys_( phys, size, align, 0, get_ram_size(), 0 ); + retain_phys = ofmem_claim_phys_( phys, size, align, 0, get_ram_size(), 1 /* reverse */ ); /* Add to the retain_phys_range list */ retained->retain_phys_range[retained->numentries].next = NULL;

1 0

[commit] r1254 - trunk/openbios-devel/libopenbios
by repository service 13 Jan '14

13 Jan '14

Author: mcayland Date: Mon Jan 13 10:46:56 2014 New Revision: 1254 URL: http://tracker.coreboot.org/trac/openbios/changeset/1254 Log: fcode_load.c: don't skip the FCode header when invoked via init-program The FCode start address was set to bypass the FCode header rather than just executing the first word which does this automatically (and may even one day verify the payload contents/length). Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> Modified: trunk/openbios-devel/libopenbios/fcode_load.c Modified: trunk/openbios-devel/libopenbios/fcode_load.c ============================================================================== --- trunk/openbios-devel/libopenbios/fcode_load.c Mon Jan 13 10:46:53 2014 (r1253) +++ trunk/openbios-devel/libopenbios/fcode_load.c Mon Jan 13 10:46:56 2014 (r1254) @@ -94,7 +94,6 @@ { /* If the payload is Fcode then we execute it immediately */ ucell address; - uint8_t fcode_header[8]; fword("load-base"); address = POP(); @@ -104,7 +103,7 @@ return; } - PUSH(address + sizeof(fcode_header)); + PUSH(address); PUSH(1); fword("byte-load"); }

1 0

[commit] r1253 - trunk/openbios-devel/forth/device
by repository service 13 Jan '14

13 Jan '14

Author: mcayland Date: Mon Jan 13 10:46:53 2014 New Revision: 1253 URL: http://tracker.coreboot.org/trac/openbios/changeset/1253 Log: other.fs: ensure the FCode function end0 is executed in immediate mode As per the IEEE1275 specification, end0 should set fcode-end to true in both FCode evaluation and execution modes. Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> Modified: trunk/openbios-devel/forth/device/other.fs Modified: trunk/openbios-devel/forth/device/other.fs ============================================================================== --- trunk/openbios-devel/forth/device/other.fs Mon Jan 13 10:46:51 2014 (r1252) +++ trunk/openbios-devel/forth/device/other.fs Mon Jan 13 10:46:53 2014 (r1253) @@ -174,7 +174,7 @@ \ Cease evaluating this FCode program. : end0 ( -- ) true fcode-end ! - ; + ; immediate \ Cease evaluating this FCode program. : end1 ( -- )

1 0

[commit] r1252 - trunk/openbios-devel/forth/device
by repository service 13 Jan '14

13 Jan '14

Author: mcayland Date: Mon Jan 13 10:46:51 2014 New Revision: 1252 URL: http://tracker.coreboot.org/trac/openbios/changeset/1252 Log: feval.fs: enable OpenBIOS to recover correctly from invalid FCode If invalid FCode has been executed, the FCode interpreter may have exited whilst still in compile mode, causing OpenBIOS internals to become corrupted. Detect this condition and switch back to immediate mode to prevent subsequent random internal errors within OpenBIOS. Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> Modified: trunk/openbios-devel/forth/device/feval.fs Modified: trunk/openbios-devel/forth/device/feval.fs ============================================================================== --- trunk/openbios-devel/forth/device/feval.fs Mon Jan 13 10:46:46 2014 (r1251) +++ trunk/openbios-devel/forth/device/feval.fs Mon Jan 13 10:46:51 2014 (r1252) @@ -49,6 +49,15 @@ , then fcode-end @ until + + \ If we've executed incorrect FCode we may have reached the end of the FCode + \ program but still be in compile mode. Make sure that if this has happened + \ then we switch back to immediate mode to prevent internal OpenBIOS errors. + tmp-comp-depth @ -1 <> if + -1 tmp-comp-depth ! + tmp-comp-buf @ @ here! + 0 state ! + then ; : byte-load ( addr xt -- )

1 0

[commit] r1251 - trunk/openbios-devel/forth/system
by repository service 13 Jan '14

13 Jan '14

Author: mcayland Date: Mon Jan 13 10:46:46 2014 New Revision: 1251 URL: http://tracker.coreboot.org/trac/openbios/changeset/1251 Log: ciface.fs: enable test-method for all architectures Currently it is only compiled for PPC, but at least *BSD on SPARC64 try to make use of it too. Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> Modified: trunk/openbios-devel/forth/system/ciface.fs Modified: trunk/openbios-devel/forth/system/ciface.fs ============================================================================== --- trunk/openbios-devel/forth/system/ciface.fs Mon Jan 13 10:46:43 2014 (r1250) +++ trunk/openbios-devel/forth/system/ciface.fs Mon Jan 13 10:46:46 2014 (r1251) @@ -315,17 +315,12 @@ outer-interpreter ; -[IFDEF] CONFIG_PPC -\ PowerPC Microprocessor CHRP binding -\ 10.5.2. Client Interface - ( cstring-method phandle -- missing ) : test-method swap dup cstrlen rot find-method 0= if -1 else drop 0 then ; -[THEN] finish-device device-end

1 0

[commit] r1250 - trunk/openbios-devel/libopenbios
by repository service 13 Jan '14

13 Jan '14

Author: mcayland Date: Mon Jan 13 10:46:43 2014 New Revision: 1250 URL: http://tracker.coreboot.org/trac/openbios/changeset/1250 Log: client.c: stop clients from being able to invoke stack underflow Clients that incorrectly requested too many return arguments for a client interface call could cause a Forth stack underflow (this was discovered due to a bug in NetBSD SPARC64). Ensure that we can never POP() beyond the original position of the stack from before the client interface call. Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> Modified: trunk/openbios-devel/libopenbios/client.c Modified: trunk/openbios-devel/libopenbios/client.c ============================================================================== --- trunk/openbios-devel/libopenbios/client.c Mon Jan 13 10:46:41 2014 (r1249) +++ trunk/openbios-devel/libopenbios/client.c Mon Jan 13 10:46:43 2014 (r1250) @@ -244,41 +244,43 @@ /* call-method, interpret */ static int -handle_calls( prom_args_t *pb ) +handle_calls(prom_args_t *pb) { - int i, dstacksave = dstackcnt; - prom_uarg_t val; + int i, j, dstacksave; + ucell val; #ifdef DEBUG_CIF printk("%s %s ([" FMT_prom_arg "] -- [" FMT_prom_arg "])\n", get_service(pb), arg2pointer(pb->args[0]), pb->nargs, pb->nret); #endif + dstacksave = dstackcnt; for (i = pb->nargs - 1; i >= 0; i--) PUSH(pb->args[i]); push_str(get_service(pb)); fword("client-call-iface"); - /* Drop the return code from client-call-iface (status is handled by the - catch result which is the first parameter below) */ - POP(); - - for (i = 0; i < pb->nret; i++) { - val = POP(); - pb->args[pb->nargs + i] = val; + /* If the catch result is non-zero, restore stack and exit */ + val = POP(); + if (val) { + dstackcnt = dstacksave; + return 0; + } - /* don't pop args if an exception occured */ - if (!i && val) - break; + j = dstackcnt; + for (i = 0; i < pb->nret; i++, j--) { + if (dstackcnt > dstacksave) { + pb->args[pb->nargs + i] = POP(); + } } #ifdef DEBUG_CIF /* useful for debug but not necessarily an error */ - if (i != pb->nret || dstackcnt != dstacksave) { + if (j != dstacksave) { printk("%s '%s': possible argument error (" FMT_prom_arg "--" FMT_prom_arg ") got %d\n", get_service(pb), arg2pointer(pb->args[0]), - pb->nargs - 2, pb->nret, i); + pb->nargs - 2, pb->nret, j - dstacksave); } printk("handle_calls return:"); @@ -293,10 +295,11 @@ } int -of_client_interface( int *params ) +of_client_interface(int *params) { prom_args_t *pb = (prom_args_t*)params; - int val, i, dstacksave; + ucell val; + int i, j, dstacksave; if (pb->nargs < 0 || pb->nret < 0 || pb->nargs + pb->nret > PROM_MAX_ARGS) @@ -317,34 +320,41 @@ push_str(get_service(pb)); fword("client-iface"); - if (val=POP()) { - dstackcnt = dstacksave; - if (val == -1) + val = POP(); + if (val) { + if (val == -1) { printk("Unimplemented service %s ([" FMT_prom_arg "] -- [" FMT_prom_arg "])\n", - get_service(pb), pb->nargs, pb->nret); + get_service(pb), pb->nargs, pb->nret); + } else { #ifdef DEBUG_CIF - else - printk("ERROR!\n"); + printk("Error calling client interface: " FMT_ucellx "\n", val); #endif + } + + dstackcnt = dstacksave; return -1; } - for (i = 0; i < pb->nret ; i++) - pb->args[pb->nargs + i] = POP(); + j = dstackcnt; + for (i = 0; i < pb->nret; i++, j--) { + if (dstackcnt > dstacksave) { + pb->args[pb->nargs + i] = POP(); + } + } - if (dstackcnt != dstacksave) { #ifdef DEBUG_CIF + if (j != dstacksave) { printk("service %s: possible argument error (%d %d)\n", - get_service(pb), i, dstackcnt - dstacksave); -#endif + get_service(pb), i, j - dstacksave); + /* Some clients request less parameters than the CIF method returns, e.g. getprop with OpenSolaris. Hence we drop any - stack parameters after issuing a warning above */ - dstackcnt = dstacksave; + stack parameters on exit after issuing a warning above */ } -#ifdef DEBUG_CIF dump_return(pb); #endif + + dstackcnt = dstacksave; return 0; }

1 0