Сначала пишется непосредственно бутлодер. Пишется как обычно, без фокусов, но линкуется подрихтованным xcl-ем. зашивается в чистый камень с установленным фузом Бутресет. Попутно решается вопрос о совместном существовании проги и лодыря -то есть переключении между ними.

вот к примеру хсl для моего лодыря. см ключи -D -Z

// File version: $Revision: 1.10 $
//
// The '_..X_' prefix is used by C-SPY as an indication that the label should
// not be displayed in the dissassembly window.
//
//////////////////////////////////////////////////////////////////////////////
/* Code (flash) segments */
-D_..X_INTVEC_SIZE=8C   /* 35 Interrupt vectors * 4 bytes each */
-D_..X_FLASH_TEND=FF    /* End of tiny flash memory */
-D_..X_FLASH_NEND=FFFF  /* End of near flash memory */
-D_..X_FLASH_END=FFFF   /* End of flash memory */
/* Internal data memory */
/* Change the two lines below to 60(BASE) and FF(TEND)
 * if you are running in "mega103 mode" */
-D_..X_SRAM_BASE=100    /* Start of ram memory */
-D_..X_SRAM_TEND=100    /* End of tiny ram memory */
-D_..X_SRAM_END=10FF    /* End of ram memory */
/* Internal EEPROM */
-D_..X_EEPROM_END=7FF   /* End of eeprom memory */
/*==================== END OF CONSTANTS =====================*/
/*
 * Modify the lines below to alter the size of the RSTACK, CSTACK and HEAP
 * segments. These need to be fine tuned to suit your specific application.
 * The '_..X_' prefix is used by C-SPY as an indication that the label should
 * not be displayed in the dissassembly window.
 */
-D_..X_CSTACK_SIZE=200  /* 512 bytes for auto variables and saved registers. */
-D_..X_RSTACK_SIZE=40   /* 64 bytes for return addresses, equivalent to 32 */
                        /* levels of calls, including interrupts. */
-D_..X_HEAP_SIZE=100    /* 256 bytes of heap. */
-D_..X_NEAR_HEAP_SIZE=100    /* 256 bytes of heap. */
/*
 * Modify these lines if you have an external SRAM connected to the system.
 *
 * Note: Remember to turn on the external data and address busses in
 *       __low_level_init if external memory is used.
 */
-D_..X_EXT_SRAM_BASE=_..X_SRAM_BASE
-D_..X_EXT_SRAM_END=_..X_SRAM_END
/*
 * Modify these lines if you have an external EPROM connected to the system.
 *
 * Note: Remember to turn on the external data and address busses in
 *       __low_level_init if external memory is used.
 */
-D_..X_EXT_EPROM_BASE=_..X_SRAM_BASE
-D_..X_EXT_EPROM_END=_..X_SRAM_END
/*
 * Modify these lines if you have an external EEPROM connected to the system.
 *
 * Note: Remember to turn on the external data and address busses in
 *       __low_level_init if external memory is used.
 */
-D_..X_EXT_EEPROM_BASE=_..X_SRAM_BASE
-D_..X_EXT_EEPROM_END=_..X_SRAM_END
/* Define CPU */
-ca90
/*
 * The following segments are located in the internal memory of
 * the ATmega64. Do not change these lines.
 */

/////////////////////////////////////////////////////////////////////////////////////////
-D_..X_SMALLBOOTSTART=F800
-D_..X_BootCodeStrt=F88C
-D_..X_BootVectEnd=F88B

/* Code memory */
-Z(CODE)INTVEC=_..X_SMALLBOOTSTART-_..X_BootVectEnd
/* Fill unused interrupt vector's with RETI */
//-H1895
//-h(CODE)0-_..X_INTVEC_SIZE
//-Z(CODE)TINY_F=_..X_BootCodeStrt-_..X_FLASH_TEND
-Z(CODE)NEAR_F,SWITCH,INITTAB,DIFUNCT=_..X_BootCodeStrt-_..X_FLASH_END
-Z(CODE)CODE=_..X_BootCodeStrt-_..X_FLASH_END
//-Z(CODE)TINY_ID=_..X_BootCodeStrt-_..X_FLASH_END
-Z(CODE)NEAR_ID=_..X_BootCodeStrt-_..X_FLASH_END
-Z(CODE)CHECKSUM#_..X_FLASH_END
/* Internal data memory */
-Z(DATA)TINY_I,TINY_Z,TINY_N=_..X_SRAM_BASE-_..X_SRAM_TEND
-Z(DATA)NEAR_I,NEAR_Z=_..X_SRAM_BASE-_..X_SRAM_END,_..X_EXT_SRAM_BASE-_..X_EXT_SRAM_END
/*
 * If external SRAM is available it is possible to place the stacks there.
 * However, the external memory is slower than the internal so moving the
 * stacks to the external memory will degrade the system performance.
 */
-Z(DATA)RSTACK+_..X_RSTACK_SIZE=_..X_SRAM_BASE-_..X_SRAM_END /* ,_..X_EXT_SRAM_BASE-_..X_EXT_SRAM_END */
-Z(DATA)CSTACK+_..X_CSTACK_SIZE=_..X_SRAM_BASE-_..X_SRAM_END /* ,_..X_EXT_SRAM_BASE-_..X_EXT_SRAM_END */
/*
 * If external SRAM is available it might be a good idea to move the
 * heap segment there, i.e. remove the _..X_SRAM_BASE-_..X_SRAM_END range.
 */
-Z(DATA)HEAP+_..X_HEAP_SIZE=_..X_SRAM_BASE-_..X_SRAM_END,_..X_EXT_SRAM_BASE-_..X_EXT_SRAM_END
-Z(DATA)IOSTREAM_N#_..X_SRAM_BASE-_..X_SRAM_END,_..X_EXT_SRAM_BASE-_..X_EXT_SRAM_END
-Z(DATA)NEAR_HEAP+_..X_NEAR_HEAP_SIZE=_..X_SRAM_BASE-_..X_SRAM_END,_..X_EXT_SRAM_BASE-_..X_EXT_SRAM_END
/* Internal eeprom memory */
-Z(XDATA)EEPROM_I,EEPROM_N=0-_..X_EEPROM_END
/*
 * The following segment definitions are only used if external memory is
 * connected to the AVR controller.
 */
/* External EPROM */
-Z(CONST)NEAR_C=_..X_EXT_EPROM_BASE-_..X_EXT_EPROM_END
/* External EEPROM */
-Z(DATA)NEAR_N=_..X_EXT_EEPROM_BASE-_..X_EXT_EEPROM_END
/* Select reduced "printf" support to reduce library size.
   See configuration section in manual concerning printf/sprintf. */
/*Dlib*/
-e_PrintfSmall=_Printf
/*Clib*/
-e_small_write=_formatted_write
-e_small_write_P=_formatted_write_P
/* Disable floating-point support in "scanf" to reduce library size.
   See configuration section in manual concerning scanf/sscanf */
/*Dlib*/
-e_ScanfSmall=_Scanf
/*Clib*/
-e_medium_read=_formatted_read
-e_medium_read_P=_formatted_read_P
/* Suppress one warning which is not relevant for this processor */
-w29
/* Code will now reside in file aout.a90 or aout.d90, unless -o is specified */
/* .d90 is the default if debug system is linked (option -r) */
/* .a90 is the default without debugging. Default format is -Fmotorola */