home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Tech Arsenal 1
/
Tech Arsenal (Arsenal Computer).ISO
/
tek-05
/
drivers1.zip
/
HEAD.ASM
< prev
next >
Wrap
Assembly Source File
|
1992-01-14
|
34KB
|
1,300 lines
include defs.asm
; Copyright, 1988-1992, Russell Nelson, Crynwr Software
; This program is free software; you can redistribute it and/or modify
; it under the terms of the GNU General Public License as published by
; the Free Software Foundation, version 1.
;
; This program is distributed in the hope that it will be useful,
; but WITHOUT ANY WARRANTY; without even the implied warranty of
; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
; GNU General Public License for more details.
;
; You should have received a copy of the GNU General Public License
; along with this program; if not, write to the Free Software
; Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
code segment word public
assume cs:code, ds:code
public phd_environ
org 2ch
phd_environ dw ?
public phd_dioa
org 80h
phd_dioa label byte
org 100h
start:
jmp start_1
extrn start_1: near
even ;put the stack on a word boundary.
;we use our dioa for a stack space. Very hard usage has shown that only
; 27 bytes were being used, so 128 should be sufficient.
our_stack label byte
extrn int_no: byte
public packet_int_no, is_at, sys_features, flagbyte
packet_int_no db ?,?,?,? ; interrupt to communicate.
is_at db 0 ; =1 if we're on an AT.
sys_features db 0 ; 2h = MC 40h = 2nd 8259
flagbyte db 0
even
functions label word
dw f_not_implemented ;0
dw f_driver_info ;1
dw f_access_type ;2
dw f_release_type ;3
dw f_send_pkt ;4
dw f_terminate ;5
dw f_get_address ;6
dw f_reset_interface ;7
dw f_stop ;8
dw f_not_implemented ;9
dw f_get_parameters ;10
dw f_not_implemented ;11
dw f_as_send_pkt ;12
dw f_drop_pkt ;13
dw f_not_implemented ;14
dw f_not_implemented ;15
dw f_not_implemented ;16
dw f_not_implemented ;17
dw f_not_implemented ;18
dw f_not_implemented ;19
dw f_set_rcv_mode ;20
dw f_get_rcv_mode ;21
dw f_set_multicast_list ;22
dw f_get_multicast_list ;23
dw f_get_statistics ;24
dw f_set_address ;25
extrn driver_class: byte
extrn driver_type: byte
extrn driver_name: byte
extrn driver_function: byte
extrn parameter_list: byte
extrn send_pkt: near
extrn as_send_pkt: near
extrn drop_pkt: near
extrn get_address: near
extrn set_address: near
extrn terminate: near
extrn reset_interface: near
extrn xmit: near
extrn recv: near
extrn recv_exiting: near
extrn etopen: near
extrn rcv_modes: word ;count of modes followed by mode handles.
extrn set_multicast_list: near
linc macro n ; inc a 32 bit integer
local a
inc n ;increment the low word
jne a ;go if not overflow
inc n+2 ;increment the high word
a:
endm
per_handle struc
in_use db 0 ;non-zero if this handle is in use.
packet_type db MAX_P_LEN dup(0);associated packet type.
packet_type_len dw 0 ;associated packet type length.
receiver dd 0 ;receiver handler.
receiver_sig db 8 dup(?) ;signature at the receiver handler.
class db ? ;interface class
per_handle ends
handles per_handle MAX_HANDLE dup(<>)
end_handles label byte
public multicast_count, multicast_addrs, multicast_broad
multicast_count dw 0 ;count of stored multicast addresses.
multicast_broad db 0ffh,0ffh,0ffh,0ffh,0ffh,0ffh ; entry for broadcast
multicast_addrs db MAX_MULTICAST*EADDR_LEN dup(?)
have_my_address db 0 ;nonzero if our address has been set.
my_address db MAX_ADDR_LEN dup(?)
my_address_len dw ?
rcv_mode_num dw 3
free_handle dw 0 ; temp, a handle not in use
found_handle dw 0 ; temp, handle for our packet
receive_ptr dd 0 ; the pkt receive service routine
public send_head, send_tail
send_head dd 0 ; head of transmit queue
send_tail dd 0 ; tail of transmit queue
statistics_list label dword
packets_in dw ?,?
packets_out dw ?,?
bytes_in dw ?,?
bytes_out dw ?,?
errors_in dw ?,?
errors_out dw ?,?
packets_dropped dw ?,? ;dropped due to no type handler.
savess dw ? ;saved during the stack swap.
savesp dw ?
regs struc ; stack offsets of incoming regs
_ES dw ?
_DS dw ?
_BP dw ?
_DI dw ?
_SI dw ?
_DX dw ?
_CX dw ?
_BX dw ?
_AX dw ?
_IP dw ?
_CS dw ?
_F dw ? ; flags, Carry flag is bit 0
regs ends
CY equ 0001h
EI equ 0200h
bytes struc ; stack offsets of incoming regs
dw ? ; es, ds, bp, di, si are 16 bits
dw ?
dw ?
dw ?
dw ?
_DL db ?
_DH db ?
_CL db ?
_CH db ?
_BL db ?
_BH db ?
_AL db ?
_AH db ?
bytes ends
public our_isr, their_isr
their_isr dd 0 ; original owner of pkt driver int
our_isr:
jmp our_isr_0 ;the required signature.
db 'PKT DRVR',0
our_isr_0:
assume ds:nothing
push ax
push bx
push cx
push dx
push si
push di
push bp
push ds
push es
cld
mov bx,cs ;set up ds.
mov ds,bx
assume ds:code
mov bp,sp ;we use bp to access the original regs.
and _F[bp],not CY ;start by clearing the carry flag.
test flagbyte,CALLED_ETOPEN ; have we initialized the card?
jnz our_isr_cont ; yes
push ax ; save lots of registers
push bx
push cx
push dx
push si
push di
push bp
push ds
push es
call etopen ; init the card
pop es ; restore lots of registers
pop ds
pop bp
pop di
pop si
pop dx
pop cx
pop bx
pop ax
mov dh,CANT_RESET ; (actually can't initialize)
jc our_isr_error
or flagbyte,CALLED_ETOPEN ; remember this fact
our_isr_cont:
mov bl,ah ;jump to the correct function.
mov bh,0
cmp bx,25 ;only twenty five functions right now.
mov dh,BAD_COMMAND ;in case we find a bad number.
ja our_isr_error
add bx,bx ;*2
call functions[bx]
jnc our_isr_return
our_isr_error:
mov _DH[bp],dh
or _F[bp],CY ;return their carry flag.
our_isr_return:
pop es
pop ds
pop bp
pop di
pop si
pop dx
pop cx
pop bx
pop ax
iret
public re_enable_interrupts
re_enable_interrupts:
; Possibly re-enable interrupts. We put this here so that other routines
; don't need to know how we put things on the stack.
test _F[bp], EI ; Were interrupts enabled on pkt driver entry?
je re_enable_interrupts_1 ; No.
sti ; Yes, re-enable interrupts now.
re_enable_interrupts_1:
ret
f_not_implemented:
mov dh,BAD_COMMAND
stc
ret
f_driver_info:
; As of 1.08, the handle is optional, so we no longer verify it.
; call verify_handle
cmp _AL[bp],0ffh ; correct calling convention?
jne f_driver_info_1 ; ne = incorrect, fail
;For enhanced PD, if they call
cmp _BX[bp],offset handles ;with a handle, give them the
;class they think it is
jb default_handle
cmp _BX[bp],offset end_handles ;otherwise default to first class
jae default_handle
mov bx, _BX[bp]
cmp [bx].in_use,0 ;if it's not in use, it's bad.
je default_handle
mov al, [bx].class
mov _CH[bp], al
jmp short got_handle
default_handle:
mov al,driver_class
mov _CH[bp],al
got_handle:
mov _BX[bp],majver ;version
mov al,driver_type
cbw
mov _DX[bp],ax
mov _CL[bp],0 ;number zero.
mov _DS[bp],ds ; point to our name in their ds:si
mov _SI[bp],offset driver_name
mov al,driver_function
mov _AL[bp],al
clc
ret
f_driver_info_1:
stc
ret
f_set_rcv_mode:
call verify_handle
cmp cx,rcv_mode_num ;are we already using that mode?
je f_set_rcv_mode_4 ;yes, no need to check anything.
mov dx,bx ;remember our handle.
mov bx,offset handles ; check that all handles are free
f_set_rcv_mode_2:
cmp bx,dx ; is this our handle?
je f_set_rcv_mode_3 ; yes, of course it's not free.
cmp [bx].in_use,0 ; is this handle free?
jne f_set_rcv_mode_1 ; ne = no, can't change
f_set_rcv_mode_3:
add bx,(size per_handle) ; next handle
cmp bx,offset end_handles ; examined all handles?
jb f_set_rcv_mode_2 ; b = no, continue examination
mov cx,_CX[bp] ;get the desired receive mode.
cmp cx,rcv_modes ;do they have this many modes?
jae f_set_rcv_mode_1 ;no - must be a bad mode for us.
mov bx,cx
add bx,bx ;we're accessing words, not bytes.
mov ax,rcv_modes[bx]+2 ;get the handler for this mode.
or ax,ax ;do they have one?
je f_set_rcv_mode_1 ;no - must be a bad mode for us.
mov rcv_mode_num,cx ;yes - remember the number and
call ax ; call it.
f_set_rcv_mode_4:
clc
ret
f_set_rcv_mode_1:
mov dh,BAD_MODE
stc
ret
f_get_rcv_mode:
call verify_handle
mov ax,rcv_mode_num ;return the current receive mode.
mov _AX[bp],ax
clc
ret
f_set_multicast_list:
mov cx,_CX[bp] ;Tell them how much room they have.
;verify that they supplied an even number of EADDR's.
mov ax,cx
xor dx,dx
mov bx,EADDR_LEN
div bx
or dx,dx ;zero remainder?
jne f_set_multicast_list_2 ;no, we don't have an even number of
; addresses.
cmp ax,MAX_MULTICAST ;is this too many?
ja f_set_multicast_list_3 ;yes - return NO_SPACE
f_set_multicast_list_1:
mov multicast_count,ax ;remember the number of addresses.
push cs
pop es
mov di,offset multicast_addrs
push ds
mov ds,_ES[bp] ; get ds:si -> new list.
mov si,_DI[bp]
push cx
rep movsb
pop cx
pop ds
mov si,offset multicast_addrs
call set_multicast_list
ret
f_set_multicast_list_2:
mov dh,BAD_ADDRESS
stc
ret
f_set_multicast_list_3:
mov dh,NO_SPACE
stc
ret
f_get_multicast_list:
mov _ES[bp],ds ;return what we have remembered.
mov _DI[bp],offset multicast_addrs
mov ax,EADDR_LEN ;multiply the count by the length.
mul multicast_count
mov _CX[bp],ax ;because they want total bytes.
clc
ret
f_get_statistics:
call verify_handle ;just in case.
mov _DS[bp],ds
mov _SI[bp],offset statistics_list
clc
ret
access_type_class:
mov dh,NO_CLASS
stc
ret
access_type_type:
mov dh,NO_TYPE
stc
ret
access_type_number:
mov dh,NO_NUMBER
stc
ret
access_type_bad:
mov dh,BAD_TYPE
stc
ret
;register caller of pkt TYPE
f_access_type:
mov bx, offset driver_class
access_type_9:
mov al, [bx] ;get the next class.
inc bx
or al,al ;end of the list?
je access_type_class ;class failed (story of my life)
cmp _AL[bp],al ;our class?
jne access_type_9 ;no, try again
access_type_1:
cmp _BX[bp],-1 ;generic type?
je access_type_2 ;yes.
mov al,driver_type
cbw
cmp _BX[bp],ax ;our type?
jne access_type_type ;no.
access_type_2:
cmp _DL[bp],0 ;generic number?
je access_type_3
cmp _DL[bp],1 ;our number?
jne access_type_number
access_type_3:
cmp _CX[bp],MAX_P_LEN ;is the type length too long?
ja access_type_bad ;yes - can't be ours.
; now we do two things--look for an open handle, and check the existing
; handles to see if they're replicating a packet type.
mov free_handle,0 ;remember no free handle yet.
mov bx,offset handles
access_type_4:
cmp [bx].in_use,0 ;is this handle in use?
je access_type_5 ;no - don't check the type.
mov al, _AL[bp] ;is this handle the same class as
cmp al, [bx].class ; they're want?
jne short access_type_6
mov es,_DS[bp] ;get a pointer to their type
mov di,_SI[bp] ; from their ds:si to our es:di
mov cx,_CX[bp] ;get the minimum of their length
; and our length. As currently
; implemented, only one receiver
; gets the packets, so we have to
; ensure that the shortest prefix
; is unique.
cmp cx,[bx].packet_type_len ;Are we less specific than they are?
jb access_type_8 ;no.
mov cx,[bx].packet_type_len ;yes - use their count.
access_type_8:
lea si,[bx].packet_type
or cx,cx ; pass-all TYPE? (zero TYPE length)
jne access_type_7 ; ne = no
mov bx,offset handles+(MAX_HANDLE-1)*(size per_handle)
jmp short access_type_5 ; put pass-all last
access_type_7:
repe cmpsb
jne short access_type_6 ;go look at the next one.
access_type_inuse:
mov dh,TYPE_INUSE ;a handle has been assigned for TYPE
stc ;and we can't assign another
ret
access_type_5: ;handle is not in use
cmp free_handle,0 ;found a free handle yet?
jne access_type_6 ;yes.
mov free_handle,bx ;remember a free handle
access_type_6:
add bx,(size per_handle) ;go to the next handle.
cmp bx,offset end_handles ;examined all handles?
jb access_type_4 ;no, continue.
mov bx,free_handle ;did we find a free handle?
or bx,bx
je access_type_space ;no - return error.
mov [bx].in_use,1 ;remember that we're using it.
mov ax,_DI[bp] ;remember the receiver type.
mov [bx].receiver.offs,ax
mov ax,_ES[bp]
mov [bx].receiver.segm,ax
push ds
mov ax,ds
mov es,ax
mov ds,_DS[bp] ;remember their type.
mov si,_SI[bp]
mov cx,_CX[bp]
mov es:[bx].packet_type_len,cx ; remember the TYPE length
lea di,[bx].packet_type
rep movsb
lds si,es:[bx].receiver ;copy the first 8 bytes
lea di,[bx].receiver_sig ; to the receiver signature.
mov cx,8/2
rep movsw
pop ds
mov al, _AL[bp]
mov [bx].class, al
mov _AX[bp],bx ;return the handle to them.
clc
ret
access_type_space:
mov dh,NO_SPACE
stc
ret
f_release_type:
call verify_handle ;mark this handle as being unused.
mov [bx].in_use,0
clc
ret
f_send_pkt:
;ds:si -> buffer, cx = length
; XXX Should re-enable interrupts here, but some drivers are broken.
; Possibly re-enable interrupts.
; test _F[bp], EI ; Were interrupts enabled on pkt driver entry?
; je f_send_pkt_1 ; No.
; sti ; Yes, re-enable interrupts now.
;f_send_pkt_1:
;following two instructions not needed because si and cx haven't been changed.
; mov si,_SI[bp]
; mov cx,_CX[bp] ; count of bytes in the packet.
linc packets_out
add bytes_out.offs,cx ;add up the received bytes.
adc bytes_out.segm,0
push ds ; set up proper ds for the buffer
mov ds,_DS[bp] ; address of buffer from caller's ds.
assume ds:nothing, es:nothing
; If -n option take Ethernet encapsulated Novell IPX packets (from BYU's
; PDSHELL) and change them to be IEEE 802.3 encapsulated.
EPROT_OFF equ EADDR_LEN*2
test cs:flagbyte,N_OPTION
jz f_send_pkt_2
cmp ds:[si].EPROT_OFF,3781h ; if not Novell (prot 8137)
jne f_send_pkt_2 ; don't tread on it
push ax ; get scratch reg
mov ax,[si].EPROT_OFF+4 ; get len
xchg ah,al
inc ax ; make even (rounding up)
and al,0feh
xchg ah,al
mov ds:[si].EPROT_OFF,ax ; save in prot field
pop ax ; restore old contents
f_send_pkt_2:
call send_pkt
pop ds
assume ds:code
ret
f_as_send_pkt:
;es:di -> iocb.
test driver_function,4 ; is this a high-performance driver?
je f_as_send_pkt_2 ; no.
; Possibly re-enable interrupts.
test _F[bp], EI ; Were interrupts enabled on pkt driver entry?
je f_as_send_pkt_1 ; No.
sti ; Yes, re-enable interrupts now.
f_as_send_pkt_1:
push ds ; set up proper ds for the buffer
lds si,es:[di].buffer ; ds:si -> buffer
assume ds:nothing
mov cx,es:[di].len ; cx = length
linc packets_out
add bytes_out.offs,cx ; add up the received bytes.
adc bytes_out.segm,0
;ds:si -> buffer, cx = length, es:di -> iocb.
call as_send_pkt
pop ds
assume ds:code
ret
f_as_send_pkt_2:
mov dh, BAD_COMMAND ; return an error.
stc
ret
f_drop_pkt:
; es:di -> iocb.
test driver_function,4 ; is this a high-performance driver?
je f_as_send_pkt_1 ; no.
push ds ; Preserve ds
mov si,offset send_head ; Get head offset
dp_loop:
mov ax,ds:[si] ; Get offset
mov dx,ds:[si+2] ; Get segment
mov bx,ax
or bx,dx ; End of list?
je dp_endlist ; Yes
cmp ax,di ; Offsets equal?
jne dp_getnext ; No
mov bx,es
cmp dx,bx ; Segments equal?
jne dp_getnext ; No
call drop_pkt ; Pass to driver
les di,es:[di].next ; Get next segment:offset
mov ds:[si],di ; Set next offset
mov ds:[si+2],es ; Set next segment
pop ds ; Restore ds
clc
ret
dp_getnext:
mov ds,dx ; Get next segment
mov si,ax ; Get next iocb offset
lea si,ds:[si].next ; Get next iocb next ptr offset
jmp dp_loop ; Try again
dp_endlist:
pop ds ; Restore ds
mov dh,BAD_IOCB ; Return error
stc ; Set carry
ret
f_terminate:
call verify_handle ; must have a handle
f_terminate_1:
mov [bx].in_use,0 ; mark handle as free
mov bx,offset handles ; check that all handles are free
f_terminate_2:
cmp [bx].in_use,0 ; is this handle free?
jne f_terminate_4 ; ne = no, so can't exit completely
add bx,(size per_handle) ; next handle
cmp bx,offset end_handles ; examined all handles?
jb f_terminate_2 ; b = no, continue examination
;
; Now disable interrupts
;
mov al,int_no
or al,al ;are they using a hardware interrupt?
je f_terminate_no_irq ;no.
call maskint
;
; Now return the interrupt to their handler.
;
mov ah,25h ;get the old interrupt into es:bx
mov al,int_no
add al,8
cmp al,8+8 ;is it a slave 8259 interrupt?
jb f_terminate_3 ;no.
add al,70h - (8+8) ;map it to the real interrupt.
f_terminate_3:
push ds
lds dx,their_recv_isr
int 21h
pop ds
f_terminate_no_irq:
call terminate ;terminate the hardware.
mov al,packet_int_no ;release our_isr.
mov ah,25h
push ds
lds dx,their_isr
int 21h
pop ds
;
; Now free our memory
;
push cs
pop es
mov ah,49h
int 21h
clc
ret
f_terminate_4:
mov dh, CANT_TERMINATE
stc
ret
f_get_address:
; call verify_handle
; mov es,_ES[bp] ; get new one
; mov di,_DI[bp] ; get pointer, es:di is ready
mov cx,_CX[bp] ;Tell them how much room they have.
cmp have_my_address,0 ;has our address been set?
jne get_address_set ;yes - go report it.
call get_address ;no, can we get the address?
jc get_address_space ;no - we must not have enough space.
mov _CX[bp],cx ;Tell them how long our address is.
clc
ret
get_address_set:
cmp cx,my_address_len ;is there enough room?
jb get_address_space ;no.
mov cx,my_address_len ;yes - get our address length.
mov _CX[bp],cx ;Tell them how long our address is.
mov si,offset my_address ;copy it into their area.
rep movsb
clc
ret
get_address_space:
mov dh,NO_SPACE
stc
ret
f_set_address:
mov bx,offset handles
mov cl,0 ;number of handles in use.
f_set_address_1:
add cl,[bx].in_use ;is this handle in use?
add bx,(size per_handle) ;go to the next handle.
cmp bx,offset end_handles
jb f_set_address_1
cmp cl,1 ;more than one handle in use?
ja f_set_address_inuse ;yes - we can't set the address
mov ds,_ES[bp] ; set new one
assume ds:nothing
mov si,_DI[bp] ; set pointer, ds:si is ready
mov cx,_CX[bp] ;Tell them how much address is being set.
call set_address
;set_address restores ds.
jc f_set_address_exit ;Did it work?
mov _CX[bp],cx ;yes - return our address length.
cmp cx,MAX_ADDR_LEN ;is it too long for us to remember?
ja f_set_address_too_long ;yes, return a too-long error.
mov ds,_ES[bp] ; set new one
mov si,_DI[bp] ; set pointer, ds:si is ready
mov ax,cs
mov es,ax
mov my_address_len,cx ;remember how long our address is.
mov di,offset my_address
rep movsb
mov have_my_address,1
mov ds,ax ;restoer ds.
assume ds:code
clc
ret
f_set_address_inuse:
mov dh,CANT_SET
stc
ret
f_set_address_too_long:
mov dh,NO_SPACE
stc
f_set_address_exit:
ret
f_reset_interface:
call verify_handle
call reset_interface
clc
ret
; Stop the packet driver doing upcalls. Also a following terminate will
; always succed (no in use handles any longer).
f_stop:
mov bx,offset handles
f_stop_2:
mov [bx].in_use,0
add bx,(size per_handle) ; next handle
cmp bx,offset end_handles
jb f_stop_2
clc
ret
f_get_parameters:
;strictly speaking, this function only works for high-performance drivers.
test driver_function,4 ;is this a high-performance driver?
jne f_get_parameters_1 ;yes.
mov dh,BAD_COMMAND ;no - return an error.
stc
ret
f_get_parameters_1:
mov _ES[bp],cs
mov _DI[bp],offset parameter_list
clc
ret
verify_handle:
;Ensure that their handle is real. If it isn't, we pop off our return
;address, and return to *their* return address with cy set.
mov bx,_BX[bp] ;get the handle they gave us
cmp bx,offset handles
jb verify_handle_bad ;no - must be bad.
cmp bx,offset end_handles
jae verify_handle_bad ;no - must be bad.
cmp [bx].in_use,0 ;if it's not in use, it's bad.
je verify_handle_bad
ret
verify_handle_bad:
mov dh,BAD_HANDLE
add sp,2 ;pop off our return address.
stc
ret
public set_recv_isr
set_recv_isr:
mov ah,35h ;get the old interrupt into es:bx
mov al,int_no ; board's interrupt vector
or al,al
je set_isr_no_irq
add al,8
cmp al,8+8 ;is it a slave 8259 interrupt?
jb set_recv_isr_1 ;no.
add al,70h - 8 - 8 ;map it to the real interrupt.
set_recv_isr_1:
int 21h
mov their_recv_isr.offs,bx ;remember the old seg:off.
mov their_recv_isr.segm,es
mov ah,25h ;now set our recv interrupt.
mov dx,offset recv_isr
int 21h
mov al,int_no ; Now enable interrupts
call unmaskint
set_isr_no_irq:
ret
public count_in_err
count_in_err:
assume ds:nothing
linc errors_in
ret
public count_out_err
count_out_err:
assume ds:nothing
linc errors_out
ret
their_recv_isr dd 0 ; original owner of board int
;
; I have had a problem with some hardware which under extreme LAN loading
; conditions will re-enter the recv_isr. Since the 8259 interrupts for
; the card are masked off, and the card's interrupt mask register is
; cleared (in 8390.asm at least) disabling the card from interrupting, this
; is clearly a hardware problem. Due to the low frequencey of occurance, and
; extreme conditions under which this happens, it is not lilely to be fixed
; in hardware any time soon, plus retrofitting of hardware in the field will
; not happen. To protect the driver from the adverse effects of this I am
; adding a simple re-entrancy trap here. - gft - 910617
;
in_recv_isr db 0 ; flag to trap re-entrancy
recv_isr:
cmp in_recv_isr, 0
je no_re_enter
iret
no_re_enter:
mov in_recv_isr, 1
; I realize this re-entrancy trap is not perfect, you could be re-entered
; anytime before the above instruction. However since the stacks have not
; been swapped re-entrancy here will only appear to be a spurious interrupt.
; - gft - 910617
; In order to achieve back-to-back packet transmissions, we handle the
; latency-critical portion of transmit interrupts first. The xmit
; interrupt routine should only start the next transmission, but do
; no other work. It may only touch ax and dx (the only register necessary
; for doing "out" instructions) unless it first pushes any other registers
; itself.
push ax
push dx
call xmit
; Now switch stacks, push remaining registers, and do remaining interrupt work.
push ds
mov ax,cs ;ds = cs.
mov ds,ax
assume ds:code
mov savesp,sp
mov savess,ss
mov ss,ax
mov sp,offset our_stack
cld
push bx
push cx
push si
push di
push bp
push es
; The following comment is wrong in that we now do a specific EOI command,
; and because we don't enable interrupts (even though we should).
; Chips & Technologies 8259 clone chip seems to be very broken. If you
; send it a Non Specific EOI command, it clears all In Service Register
; bits instead of just the one with the highest priority (as the Intel
; chip does and clones should do). This bug causes our interrupt
; routine to be reentered if: 1. we reenable processor interrupts;
; 2. we reenable device interrupts; 3. a timer or other higher priority
; device interrupt now comes in; 4. the new interrupting device uses
; a Non Specific EOI; 5. our device interrupts again. Because of
; this bug, we now completely mask our interrupts around the call
; to "recv", the real device interrupt handler. This allows us
; to send an EOI instruction to the 8259 early, before we actually
; reenable device interrupts. Since the interrupt is masked, we
; are still guaranteed not to get another interrupt from our device
; until the interrupt handler returns. This has another benefit:
; we now no longer prevent other devices from interrupting while our
; interrupt handler is running. This is especially useful if we have
; other (multiple) packet drivers trying to do low-latency transmits.
mov al,int_no ; Disable further device interrupts
call maskint
; The following is from Bill Rust, <wjr@ftp.com>
; this code dismisses the interrupt at the 8259. if the interrupt number
; is > 8 then it requires fondling two PICs instead of just one.
mov al, int_no ; get hardware int #
cmp al, 8 ; see if its on secondary PIC
jg recv_isr_4
add al, 60h ; make specific EOI dismissal
out 20h, al
jmp recv_isr_3 ; all done
recv_isr_4:
add al,60h - 8 ; make specific EOI (# between 9 & 15).
out 0a0h,al ; Secondary 8259 (PC/AT only)
mov al,62h ; Acknowledge on primary 8259.
out 20h,al
recv_isr_3:
; sti ; Interrupts are now completely safe
call recv
cli ;interrupts *must* be off between
;here and the stack restore, because
;if we have one of our interrupts
;pending, we would trash our stack.
;
; According to Novell IMSP Technical Memo #2, Lost Interrupts and NetWare,
; some LAN controllers can glitch the interrupt line when the interrupt
; mask register on the LAN controller is enabled to allow the LAN controller
; to interrupt. This can cause spurious/lost interrupt problems, especially
; on some 486 processors which latch the int line to the processor. To prevent
; glitches from being propogated throug, move the call to recv_exiting (re-
; enables interrupts on the LAN card) to before when the 8259 interrupts are
; unmasked. - gft - 910617
call recv_exiting
mov al,int_no ; Now reenable device interrupts
call unmaskint
pop es
pop bp
pop di
pop si
pop cx
pop bx
mov ss,savess
mov sp,savesp
; move the next call up to above the call unmaskint - gft - 910617
; call recv_exiting ;this routine can enable interrupts.
; DDP - This is a BIG mistake. This routine SHOULD NOT enable interrupts.
; doing so can cause interrupt recursion and blow your stack.
; Processor interrupts SHOULD NOT be enabled after enabling device
; interrupts until after the "iret". You will lose atleast 12 bytes
; on the stack for each recursion.
pop ds
assume ds:nothing
pop dx
pop ax
mov in_recv_isr, 0 ; clear the re-entrancy flag - gft - 901617
iret
public maskint
maskint:
or al,al ;are they using a hardware interrupt?
je maskint_1 ;no, don't mask off the timer!
assume ds:code
mov dx,21h ;assume the master 8259.
cmp al,8 ;using the slave 8259 on an AT?
jb mask_not_irq2
mov dx,0a1h ;go disable it on slave 8259
sub al,8
mask_not_irq2:
mov cl,al
in al,dx ;disable them on the correct 8259.
mov ah,1 ;set the bit.
shl ah,cl
or al,ah
;
; 500ns Stall required here, per INTEL documentation for eisa machines
; - gft - 910617
;
push ax
in al,61h ; 1.5 - 3 uS should be plenty
in al,61h
in al,61h
pop ax
out dx,al
maskint_1:
ret
public unmaskint
unmaskint:
assume ds:code
mov dx,21h ;assume the master 8259.
mov cl,al
cmp cl,8 ;using the slave 8259 on an AT?
jb unmask_not_irq2 ;no
in al,dx ;get master mask
and al,not (1 shl 2) ; and clear slave cascade bit in mask
out dx,al ;set new master mask (enable slave int)
;
; 500ns Stall required here, per INTEL documentation for eisa machines
; - gft - 910617
;
push ax
in al,61h ; 1.5 - 3 uS should be plenty
in al,61h
in al,61h
pop ax
mov dx,0a1h ;go enable int on slave 8259
sub cl,8
unmask_not_irq2:
in al,dx ;enable interrupts on the correct 8259.
mov ah,1 ;clear the bit.
shl ah,cl
not ah
and al,ah
;
; 500ns Stall required here, per INTEL documentation for eisa machines
; - gft - 910617
;
push ax
in al,61h ; 1.5 - 3 uS should be plenty
in al,61h
in al,61h
pop ax
out dx,al
ret
public recv_find
recv_find:
;called when we want to determine what to do with a received packet.
;enter with cx = packet length, es:di -> packet type, dl = packet class.
;exit with es:di = 0 if the packet is not desired, or es:di -> packet buffer
; to be filled by the driver.
assume ds:code, es:nothing
push cx
; If -n option take IEEE 802.3 encapsulated packets that could be Novell IPX
; and make them Ethernet encapsulated Novell IPX packets (for PDSHELL).
test flagbyte,N_OPTION
jz not_n_op
; Make IEEE 802.3-like packets that could be Novell IPX into BlueBook class
; Novell type 8137 packets.
cmp dl,IEEE8023 ;Is this an IEEE 802.3 packet?
jne recv_not_802_3 ;no
cmp word ptr es:[di],0ffffh ;if this word not ffff
jne recv_not_8137 ; then not Novell
sub di,2 ; back it up to the 8137 word.
mov es:[di],3781h ; fake as Novell protocol (8137)
mov dl,BLUEBOOK
jmp short recv_not_8137
recv_not_802_3:
; Convert incoming Ethernet type 8137 IPX packets to type 8138, as with -n in
; effect we can't send type 8137, and it will only confuse Netware.
cmp dl,BLUEBOOK ;Is this a BLUEBOOK packet?
jne recv_not_8137 ;no, don't change it.
cmp word ptr es:[di],3781h ;Is it an 8137 packet?
jne recv_not_8137 ;no, don't change it.
mov es:[di],word ptr 3881h ;yes, mung it slightly.
recv_not_8137:
not_n_op:
mov bx,offset handles
recv_find_1:
cmp [bx].in_use,0 ;is this handle in use?
je recv_find_2 ;no - don't check the type.
mov ax,[bx].receiver.offs ;do they have a receiver?
or ax,[bx].receiver.segm
je recv_find_2 ;no - they're not serious about it.
mov cx,[bx].packet_type_len ;compare the packets.
lea si,[bx].packet_type
jcxz recv_find_3 ;if cx is zero, they want them all.
cmp [bx].class, dl ;is this the right class?
jne recv_find_2 ;no- don't bother
push di
repe cmpsb
pop di
je recv_find_3 ;we've got it!
recv_find_2:
add bx,(size per_handle) ;go to the next handle.
cmp bx,offset end_handles
jb recv_find_1
linc packets_dropped
pop cx ;we didn't find it -- discard it.
recv_find_5:
xor di,di ;"return" a null pointer.
mov es,di
ret
recv_find_3:
pop cx ; the packet_length
linc packets_in
add bytes_in.offs,cx ;add up the received bytes.
adc bytes_in.segm,0
les di,[bx].receiver ;remember the receiver upcall.
mov receive_ptr.offs,di
mov receive_ptr.segm,es
test flagbyte,W_OPTION ;did they select the Windows option?
je recv_find_6 ;no, don't check for the upcall.
; does the receiver signature match whats currently in memory? if not,
; jump to fake return
push si
push cx
lea si,[bx].receiver_sig
mov cx,8/2
repe cmpsw
pop cx
pop si
jne recv_find_5
recv_find_6:
mov found_handle,bx ;remember what our handle was.
mov ax,0 ;allocate request.
stc ;with stc, flags must be an odd number
push ax ; save a number that cant be flags
pushf ;save flags in case iret used.
call receive_ptr ;ask the client for a buffer.
; on return, flags should be at top of stack. if an IRET has been used,
; then 0 will be at the top of the stack
pop bx
cmp bx,0
je recv_find_4 ;0 is at top of stack
add sp,2
recv_find_4:
ret
public recv_copy
recv_copy:
;called after we have copied the packet into the buffer.
;enter with ds:si ->the packet, cx = length of the packet.
;preserve bx.
assume ds:nothing, es:nothing
push bx
mov bx,found_handle
mov ax,1 ;store request.
clc ;with clc, flags must be an even number
push ax ; save a number that can't be flags
pushf ;save flags incase iret used.
call receive_ptr ;ask the client for a buffer.
pop bx
cmp bx,1 ;if this is a 1, IRET was used.
je recv_copy_1
pop bx
recv_copy_1:
pop bx
ret
public send_queue
send_queue:
; Queue an iocb.
; Enter with es:di -> iocb, interrupts disabled.
; Destroys ds:si.
assume ds:nothing, es:nothing
mov es:[di].next.offs,0 ; Zero next offset
mov es:[di].next.segm,0 ; Zero next segment
mov si,send_head.offs ; Queue empty?
or si,send_head.segm
jnz sq_notempty ; No
mov send_head.offs,di ; Set head offset
mov send_head.segm,es ; Set head segment
jmp sq_settail
sq_notempty: ; Queue is not empty
lds si,send_tail ; Get tail segment:offset
mov ds:[si].next.offs,di ; Set next offset
mov ds:[si].next.segm,es ; Set next segment
sq_settail:
mov send_tail.offs,di ; Set tail offset
mov send_tail.segm,es ; Set tail segment
ret
public send_dequeue
send_dequeue:
; Dequeue an iocb and possibly call its upcall.
; Enter with device or processor interrupts disabled, ah = return code.
; Exits with es:di -> iocb; destroys ds:si, ax, bx, cx, dx, bp.
assume ds:nothing, es:nothing
les di,send_head ; Get head segment:offset
lds si,es:[di].next ; Get next segment:offset
mov send_head.offs, si ; Set head offset
mov send_head.segm, ds ; Set head segment
or es:flags[di], DONE ; Mark done
mov es:ret_code[di], ah ; Set retcode
test es:[di].flags,CALLME ; Does he want an upcall?
je send_dequeue_1 ; No.
push es ; Push iocb segment
push di ; and offset
clc ; Clear carry.
mov ax,1 ; Push a number that cant be flags.
push ax
pushf ; Save flags in case iret used.
call es:[di].upcall ; Call the client.
pop ax ; Pop first word.
cmp ax,1 ; If this is a 1, IRET was used.
je send_dequeue_2 ; Far return used.
add sp,2 ; Pop flags.
send_dequeue_2:
pop di ; Pop iocb segment
pop es ; and offset
send_dequeue_1:
ret
code ends
end start
