2 Debugging on Linux for s/390 & z/Architecture
4 Denis Joseph Barrow (djbarrow@de.ibm.com,barrow_dj@yahoo.com)
5 Copyright (C) 2000-2001 IBM Deutschland Entwicklung GmbH, IBM Corporation
6 Best viewed with fixed width fonts
10 This document is intended to give a good overview of how to debug Linux for
11 s/390 and z/Architecture. It is not intended as a complete reference and not a
12 tutorial on the fundamentals of C & assembly. It doesn't go into
13 390 IO in any detail. It is intended to complement the documents in the
14 reference section below & any other worthwhile references you get.
16 It is intended like the Enterprise Systems Architecture/390 Reference Summary
17 to be printed out & used as a quick cheat sheet self help style reference when
23 Address Spaces on Intel Linux
24 Address Spaces on Linux for s/390 & z/Architecture
25 The Linux for s/390 & z/Architecture Kernel Task Structure
26 Register Usage & Stackframes on Linux for s/390 & z/Architecture
27 A sample program with comments
28 Compiling programs for debugging on Linux for s/390 & z/Architecture
30 s/390 & z/Architecture IO Overview
31 Debugging IO on s/390 & z/Architecture under VM
32 GDB on s/390 & z/Architecture
33 Stack chaining in gdb by hand
44 The current architectures have the following registers.
46 16 General propose registers, 32 bit on s/390 and 64 bit on z/Architecture,
47 r0-r15 (or gpr0-gpr15), used for arithmetic and addressing.
49 16 Control registers, 32 bit on s/390 and 64 bit on z/Architecture, cr0-cr15,
50 kernel usage only, used for memory management, interrupt control, debugging
53 16 Access registers (ar0-ar15), 32 bit on both s/390 and z/Architecture,
54 normally not used by normal programs but potentially could be used as
55 temporary storage. These registers have a 1:1 association with general
56 purpose registers and are designed to be used in the so-called access
57 register mode to select different address spaces.
58 Access register 0 (and access register 1 on z/Architecture, which needs a
59 64 bit pointer) is currently used by the pthread library as a pointer to
60 the current running threads private area.
62 16 64 bit floating point registers (fp0-fp15 ) IEEE & HFP floating
63 point format compliant on G5 upwards & a Floating point control reg (FPC)
64 4 64 bit registers (fp0,fp2,fp4 & fp6) HFP only on older machines.
66 Linux (currently) always uses IEEE & emulates G5 IEEE format on older machines,
67 ( provided the kernel is configured for this ).
70 The PSW is the most important register on the machine it
71 is 64 bit on s/390 & 128 bit on z/Architecture & serves the roles of
72 a program counter (pc), condition code register,memory space designator.
73 In IBM standard notation I am counting bit 0 as the MSB.
74 It has several advantages over a normal program counter
75 in that you can change address translation & program counter
76 in a single instruction. To change address translation,
77 e.g. switching address translation off requires that you
78 have a logical=physical mapping for the address you are
83 0 0 Reserved ( must be 0 ) otherwise specification exception occurs.
85 1 1 Program Event Recording 1 PER enabled,
86 PER is used to facilitate debugging e.g. single stepping.
88 2-4 2-4 Reserved ( must be 0 ).
90 5 5 Dynamic address translation 1=DAT on.
92 6 6 Input/Output interrupt Mask
94 7 7 External interrupt Mask used primarily for interprocessor
95 signalling and clock interrupts.
97 8-11 8-11 PSW Key used for complex memory protection mechanism
98 (not used under linux)
100 12 12 1 on s/390 0 on z/Architecture
102 13 13 Machine Check Mask 1=enable machine check interrupts
104 14 14 Wait State. Set this to 1 to stop the processor except for
105 interrupts and give time to other LPARS. Used in CPU idle in
106 the kernel to increase overall usage of processor resources.
108 15 15 Problem state ( if set to 1 certain instructions are disabled )
109 all linux user programs run with this bit 1
110 ( useful info for debugging under VM ).
112 16-17 16-17 Address Space Control
114 00 Primary Space Mode:
115 The register CR1 contains the primary address-space control ele-
116 ment (PASCE), which points to the primary space region/segment
119 01 Access register mode
121 10 Secondary Space Mode:
122 The register CR7 contains the secondary address-space control
123 element (SASCE), which points to the secondary space region or
124 segment table origin.
127 The register CR13 contains the home space address-space control
128 element (HASCE), which points to the home space region/segment
131 See "Address Spaces on Linux for s/390 & z/Architecture" below
132 for more information about address space usage in Linux.
134 18-19 18-19 Condition codes (CC)
136 20 20 Fixed point overflow mask if 1=FPU exceptions for this event
139 21 21 Decimal overflow mask if 1=FPU exceptions for this event occur
142 22 22 Exponent underflow mask if 1=FPU exceptions for this event occur
145 23 23 Significance Mask if 1=FPU exceptions for this event occur
148 24-31 24-30 Reserved Must be 0.
150 31 Extended Addressing Mode
151 32 Basic Addressing Mode
152 Used to set addressing mode
158 32 1=31 bit addressing mode 0=24 bit addressing mode (for backward
159 compatibility), linux always runs with this bit set to 1
161 33-64 Instruction address.
162 33-63 Reserved must be 0
164 In 24 bits mode bits 64-103=0 bits 104-127 Address
165 In 31 bits mode bits 64-96=0 bits 97-127 Address
166 Note: unlike 31 bit mode on s/390 bit 96 must be zero
167 when loading the address with LPSWE otherwise a
168 specification exception occurs, LPSW is fully backward
174 This per cpu memory area is too intimately tied to the processor not to mention.
175 It exists between the real addresses 0-4096 on s/390 and between 0-8192 on
176 z/Architecture and is exchanged with one page on s/390 or two pages on
177 z/Architecture in absolute storage by the set prefix instruction during Linux
179 This page is mapped to a different prefix for each processor in an SMP
180 configuration (assuming the OS designer is sane of course).
181 Bytes 0-512 (200 hex) on s/390 and 0-512, 4096-4544, 4604-5119 currently on
182 z/Architecture are used by the processor itself for holding such information
183 as exception indications and entry points for exceptions.
184 Bytes after 0xc00 hex are used by linux for per processor globals on s/390 and
185 z/Architecture (there is a gap on z/Architecture currently between 0xc00 and
186 0x1000, too, which is used by Linux).
187 The closest thing to this on traditional architectures is the interrupt
188 vector table. This is a good thing & does simplify some of the kernel coding
189 however it means that we now cannot catch stray NULL pointers in the
190 kernel without hard coded checks.
194 Address Spaces on Intel Linux
195 =============================
197 The traditional Intel Linux is approximately mapped as follows forgive
199 0xFFFFFFFF 4GB Himem *****************
203 ***************** ****************
204 User Space Himem * User Stack * * *
205 (typically 0xC0000000 3GB ) ***************** * *
206 * Shared Libs * * Next Process *
207 ***************** * to *
213 0x00000000 ***************** ****************
215 Now it is easy to see that on Intel it is quite easy to recognise a kernel
216 address as being one greater than user space himem (in this case 0xC0000000),
217 and addresses of less than this are the ones in the current running program on
218 this processor (if an smp box).
219 If using the virtual machine ( VM ) as a debugger it is quite difficult to
220 know which user process is running as the address space you are looking at
221 could be from any process in the run queue.
223 The limitation of Intels addressing technique is that the linux
224 kernel uses a very simple real address to virtual addressing technique
225 of Real Address=Virtual Address-User Space Himem.
226 This means that on Intel the kernel linux can typically only address
227 Himem=0xFFFFFFFF-0xC0000000=1GB & this is all the RAM these machines
229 They can lower User Himem to 2GB or lower & thus be
230 able to use 2GB of RAM however this shrinks the maximum size
231 of User Space from 3GB to 2GB they have a no win limit of 4GB unless
235 On 390 our limitations & strengths make us slightly different.
236 For backward compatibility we are only allowed use 31 bits (2GB)
237 of our 32 bit addresses, however, we use entirely separate address
238 spaces for the user & kernel.
240 This means we can support 2GB of non Extended RAM on s/390, & more
241 with the Extended memory management swap device &
242 currently 4TB of physical memory currently on z/Architecture.
245 Address Spaces on Linux for s/390 & z/Architecture
246 ==================================================
248 Our addressing scheme is basically as follows:
250 Primary Space Home Space
251 Himem 0x7fffffff 2GB on s/390 ***************** ****************
252 currently 0x3ffffffffff (2^42)-1 * User Stack * * *
253 on z/Architecture. ***************** * *
255 ***************** * *
261 0x00000000 ***************** ****************
263 This also means that we need to look at the PSW problem state bit and the
264 addressing mode to decide whether we are looking at user or kernel space.
266 User space runs in primary address mode (or access register mode within
269 The kernel usually also runs in home space mode, however when accessing
270 user space the kernel switches to primary or secondary address mode if
271 the mvcos instruction is not available or if a compare-and-swap (futex)
272 instruction on a user space address is performed.
274 When also looking at the ASCE control registers, this means:
277 - runs in primary or access register mode
278 - cr1 contains the user asce
279 - cr7 contains the user asce
280 - cr13 contains the kernel asce
283 - runs in home space mode
284 - cr1 contains the user or kernel asce
285 -> the kernel asce is loaded when a uaccess requires primary or
286 secondary address mode
287 - cr7 contains the user or kernel asce, (changed with set_fs())
288 - cr13 contains the kernel asce
290 In case of uaccess the kernel changes to:
291 - primary space mode in case of a uaccess (copy_to_user) and uses
292 e.g. the mvcp instruction to access user space. However the kernel
293 will stay in home space mode if the mvcos instruction is available
294 - secondary space mode in case of futex atomic operations, so that the
295 instructions come from primary address space and data from secondary
298 In case of KVM, the kernel runs in home space mode, but cr1 gets switched
299 to contain the gmap asce before the SIE instruction gets executed. When
300 the SIE instruction is finished, cr1 will be switched back to contain the
304 Virtual Addresses on s/390 & z/Architecture
305 ===========================================
307 A virtual address on s/390 is made up of 3 parts
308 The SX (segment index, roughly corresponding to the PGD & PMD in Linux
309 terminology) being bits 1-11.
310 The PX (page index, corresponding to the page table entry (pte) in Linux
311 terminology) being bits 12-19.
312 The remaining bits BX (the byte index are the offset in the page )
315 On z/Architecture in linux we currently make up an address from 4 parts.
316 The region index bits (RX) 0-32 we currently use bits 22-32
317 The segment index (SX) being bits 33-43
318 The page index (PX) being bits 44-51
319 The byte index (BX) being bits 52-63
322 1) s/390 has no PMD so the PMD is really the PGD also.
323 A lot of this stuff is defined in pgtable.h.
325 2) Also seeing as s/390's page indexes are only 1k in size
326 (bits 12-19 x 4 bytes per pte ) we use 1 ( page 4k )
327 to make the best use of memory by updating 4 segment indices
328 entries each time we mess with a PMD & use offsets
329 0,1024,2048 & 3072 in this page as for our segment indexes.
330 On z/Architecture our page indexes are now 2k in size
331 ( bits 12-19 x 8 bytes per pte ) we do a similar trick
332 but only mess with 2 segment indices each time we mess with
335 3) As z/Architecture supports up to a massive 5-level page table lookup we
336 can only use 3 currently on Linux ( as this is all the generic kernel
337 currently supports ) however this may change in future
338 this allows us to access ( according to my sums )
339 4TB of virtual storage per process i.e.
340 4096*512(PTES)*1024(PMDS)*2048(PGD) = 4398046511104 bytes,
341 enough for another 2 or 3 of years I think :-).
342 to do this we use a region-third-table designation type in
343 our address space control registers.
346 The Linux for s/390 & z/Architecture Kernel Task Structure
347 ==========================================================
348 Each process/thread under Linux for S390 has its own kernel task_struct
349 defined in linux/include/linux/sched.h
350 The S390 on initialisation & resuming of a process on a cpu sets
351 the __LC_KERNEL_STACK variable in the spare prefix area for this cpu
352 (which we use for per-processor globals).
354 The kernel stack pointer is intimately tied with the task structure for
355 each processor as follows.
358 ************************
359 * 1 page kernel stack *
361 ************************
362 * 1 page task_struct *
364 8K aligned ************************
367 ************************
368 * 2 page kernel stack *
370 ************************
371 * 2 page task_struct *
373 16K aligned ************************
375 What this means is that we don't need to dedicate any register or global
376 variable to point to the current running process & can retrieve it with the
377 following very simple construct for s/390 & one very similar for z/Architecture.
379 static inline struct task_struct * get_current(void)
381 struct task_struct *current;
382 __asm__("lhi %0,-8192\n\t"
388 i.e. just anding the current kernel stack pointer with the mask -8192.
389 Thankfully because Linux doesn't have support for nested IO interrupts
390 & our devices have large buffers can survive interrupts being shut for
391 short amounts of time we don't need a separate stack for interrupts.
396 Register Usage & Stackframes on Linux for s/390 & z/Architecture
397 =================================================================
400 This is the code that gcc produces at the top & the bottom of
401 each function. It usually is fairly consistent & similar from
402 function to function & if you know its layout you can probably
403 make some headway in finding the ultimate cause of a problem
404 after a crash without a source level debugger.
406 Note: To follow stackframes requires a knowledge of C or Pascal &
407 limited knowledge of one assembly language.
409 It should be noted that there are some differences between the
410 s/390 and z/Architecture stack layouts as the z/Architecture stack layout
411 didn't have to maintain compatibility with older linkage formats.
416 This is a built in compiler function for runtime allocation
417 of extra space on the callers stack which is obviously freed
418 up on function exit ( e.g. the caller may choose to allocate nothing
419 of a buffer of 4k if required for temporary purposes ), it generates
420 very efficient code ( a few cycles ) when compared to alternatives
423 automatics: These are local variables on the stack,
424 i.e they aren't in registers & they aren't static.
427 This is a pointer to the stack pointer before entering a
428 framed functions ( see frameless function ) prologue got by
429 dereferencing the address of the current stack pointer,
430 i.e. got by accessing the 32 bit value at the stack pointers
434 This is a pointer to the back of the literal pool which
435 is an area just behind each procedure used to store constants
438 call-clobbered: The caller probably needs to save these registers if there
439 is something of value in them, on the stack or elsewhere before making a
440 call to another procedure so that it can restore it later.
443 The code generated by the compiler to return to the caller.
446 A frameless function in Linux for s390 & z/Architecture is one which doesn't
447 need more than the register save area (96 bytes on s/390, 160 on z/Architecture)
448 given to it by the caller.
449 A frameless function never:
450 1) Sets up a back chain.
452 3) Calls other normal functions
456 This is a pointer to the global-offset-table in ELF
457 ( Executable Linkable Format, Linux'es most common executable format ),
458 all globals & shared library objects are found using this pointer.
461 ELF shared libraries are typically only loaded when routines in the shared
462 library are actually first called at runtime. This is lazy binding.
464 procedure-linkage-table
465 This is a table found from the GOT which contains pointers to routines
466 in other shared libraries which can't be called to by easier means.
469 The code generated by the compiler to set up the stack frame.
472 This is extra area allocated on the stack of the calling function if the
473 parameters for the callee's cannot all be put in registers, the same
474 area can be reused by each function the caller calls.
477 A COFF executable format based concept of a procedure reference
478 actually being 8 bytes or more as opposed to a simple pointer to the routine.
479 This is typically defined as follows
480 Routine Descriptor offset 0=Pointer to Function
481 Routine Descriptor offset 4=Pointer to Table of Contents
482 The table of contents/TOC is roughly equivalent to a GOT pointer.
483 & it means that shared libraries etc. can be shared between several
484 environments each with their own TOC.
487 static-chain: This is used in nested functions a concept adopted from pascal
488 by gcc not used in ansi C or C++ ( although quite useful ), basically it
489 is a pointer used to reference local variables of enclosing functions.
490 You might come across this stuff once or twice in your lifetime.
493 The function below should return 11 though gcc may get upset & toss warnings
494 about unused variables.
507 s/390 & z/Architecture Register usage
508 =====================================
509 r0 used by syscalls/assembly call-clobbered
510 r1 used by syscalls/assembly call-clobbered
511 r2 argument 0 / return value 0 call-clobbered
512 r3 argument 1 / return value 1 (if long long) call-clobbered
513 r4 argument 2 call-clobbered
514 r5 argument 3 call-clobbered
516 r7 pointer-to arguments 5 to ... saved
519 r10 static-chain ( if nested function ) saved
520 r11 frame-pointer ( if function used alloca ) saved
521 r12 got-pointer saved
522 r13 base-pointer saved
523 r14 return-address saved
524 r15 stack-pointer saved
526 f0 argument 0 / return value ( float/double ) call-clobbered
527 f2 argument 1 call-clobbered
528 f4 z/Architecture argument 2 saved
529 f6 z/Architecture argument 3 saved
530 The remaining floating points
531 f1,f3,f5 f7-f15 are call-clobbered.
535 1) The only requirement is that registers which are used
536 by the callee are saved, e.g. the compiler is perfectly
537 capable of using r11 for purposes other than a frame a
538 frame pointer if a frame pointer is not needed.
539 2) In functions with variable arguments e.g. printf the calling procedure
540 is identical to one without variable arguments & the same number of
541 parameters. However, the prologue of this function is somewhat more
542 hairy owing to it having to move these parameters to the stack to
543 get va_start, va_arg & va_end to work.
544 3) Access registers are currently unused by gcc but are used in
545 the kernel. Possibilities exist to use them at the moment for
546 temporary storage but it isn't recommended.
547 4) Only 4 of the floating point registers are used for
548 parameter passing as older machines such as G3 only have only 4
549 & it keeps the stack frame compatible with other compilers.
550 However with IEEE floating point emulation under linux on the
551 older machines you are free to use the other 12.
552 5) A long long or double parameter cannot be have the
553 first 4 bytes in a register & the second four bytes in the
554 outgoing args area. It must be purely in the outgoing args
555 area if crossing this boundary.
556 6) Floating point parameters are mixed with outgoing args
557 on the outgoing args area in the order the are passed in as parameters.
558 7) Floating point arguments 2 & 3 are saved in the outgoing args area for
565 0 0 back chain ( a 0 here signifies end of back chain )
566 4 8 eos ( end of stack, not used on Linux for S390 used in other linkage formats )
567 8 16 glue used in other s/390 linkage formats for saved routine descriptors etc.
568 12 24 glue used in other s/390 linkage formats for saved routine descriptors etc.
571 24 48 saved r6 of caller function
572 28 56 saved r7 of caller function
573 32 64 saved r8 of caller function
574 36 72 saved r9 of caller function
575 40 80 saved r10 of caller function
576 44 88 saved r11 of caller function
577 48 96 saved r12 of caller function
578 52 104 saved r13 of caller function
579 56 112 saved r14 of caller function
580 60 120 saved r15 of caller function
581 64 128 saved f4 of caller function
582 72 132 saved f6 of caller function
584 96 160 outgoing args passed from caller to callee
585 96+x 160+x possible stack alignment ( 8 bytes desirable )
586 96+x+y 160+x+y alloca space of caller ( if used )
587 96+x+y+z 160+x+y+z automatics of caller ( if used )
590 A sample program with comments.
591 ===============================
593 Comments on the function test
594 -----------------------------
595 1) It didn't need to set up a pointer to the constant pool gpr13 as it is not
597 2) This is a frameless function & no stack is bought.
598 3) The compiler was clever enough to recognise that it could return the
599 value in r2 as well as use it for the passed in parameter ( :-) ).
600 4) The basr ( branch relative & save ) trick works as follows the instruction
601 has a special case with r0,r0 with some instruction operands is understood as
602 the literal value 0, some risc architectures also do this ). So now
603 we are branching to the next address & the address new program counter is
604 in r13,so now we subtract the size of the function prologue we have executed
605 + the size of the literal pool to get to the top of the literal pool
606 0040037c int test(int b)
607 { # Function prologue below
608 40037c: 90 de f0 34 stm %r13,%r14,52(%r15) # Save registers r13 & r14
609 400380: 0d d0 basr %r13,%r0 # Set up pointer to constant pool using
610 400382: a7 da ff fa ahi %r13,-6 # basr trick
613 400386: a7 2a 00 05 ahi %r2,5 # add 5 to r2
615 # Function epilogue below
616 40038a: 98 de f0 34 lm %r13,%r14,52(%r15) # restore registers r13 & 14
617 40038e: 07 fe br %r14 # return
620 Comments on the function main
621 -----------------------------
622 1) The compiler did this function optimally ( 8-) )
624 Literal pool for main.
625 400390: ff ff ff ec .long 0xffffffec
626 main(int argc,char *argv[])
627 { # Function prologue below
628 400394: 90 bf f0 2c stm %r11,%r15,44(%r15) # Save necessary registers
629 400398: 18 0f lr %r0,%r15 # copy stack pointer to r0
630 40039a: a7 fa ff a0 ahi %r15,-96 # Make area for callee saving
631 40039e: 0d d0 basr %r13,%r0 # Set up r13 to point to
632 4003a0: a7 da ff f0 ahi %r13,-16 # literal pool
633 4003a4: 50 00 f0 00 st %r0,0(%r15) # Save backchain
635 return(test(5)); # Main Program Below
636 4003a8: 58 e0 d0 00 l %r14,0(%r13) # load relative address of test from
638 4003ac: a7 28 00 05 lhi %r2,5 # Set first parameter to 5
639 4003b0: 4d ee d0 00 bas %r14,0(%r14,%r13) # jump to test setting r14 as return
640 # address using branch & save instruction.
642 # Function Epilogue below
643 4003b4: 98 bf f0 8c lm %r11,%r15,140(%r15)# Restore necessary registers.
644 4003b8: 07 fe br %r14 # return to do program exit
651 main(int argc,char *argv[])
653 4004fc: 90 7f f0 1c stm %r7,%r15,28(%r15)
654 400500: a7 d5 00 04 bras %r13,400508 <main+0xc>
655 400504: 00 40 04 f4 .long 0x004004f4
656 # compiler now puts constant pool in code to so it saves an instruction
657 400508: 18 0f lr %r0,%r15
658 40050a: a7 fa ff a0 ahi %r15,-96
659 40050e: 50 00 f0 00 st %r0,0(%r15)
661 400512: 58 10 d0 00 l %r1,0(%r13)
662 400516: a7 28 00 05 lhi %r2,5
663 40051a: 0d e1 basr %r14,%r1
664 # compiler adds 1 extra instruction to epilogue this is done to
665 # avoid processor pipeline stalls owing to data dependencies on g5 &
666 # above as register 14 in the old code was needed directly after being loaded
667 # by the lm %r11,%r15,140(%r15) for the br %14.
668 40051c: 58 40 f0 98 l %r4,152(%r15)
669 400520: 98 7f f0 7c lm %r7,%r15,124(%r15)
674 Hartmut ( our compiler developer ) also has been threatening to take out the
675 stack backchain in optimised code as this also causes pipeline stalls, you
678 64 bit z/Architecture code disassembly
679 --------------------------------------
681 If you understand the stuff above you'll understand the stuff
682 below too so I'll avoid repeating myself & just say that
683 some of the instructions have g's on the end of them to indicate
684 they are 64 bit & the stack offsets are a bigger,
685 the only other difference you'll find between 32 & 64 bit is that
686 we now use f4 & f6 for floating point arguments on 64 bit.
687 00000000800005b0 <test>:
691 800005b0: a7 2a 00 05 ahi %r2,5
692 800005b4: b9 14 00 22 lgfr %r2,%r2 # downcast to integer
693 800005b8: 07 fe br %r14
694 800005ba: 07 07 bcr 0,%r7
699 00000000800005bc <main>:
700 main(int argc,char *argv[])
702 800005bc: eb bf f0 58 00 24 stmg %r11,%r15,88(%r15)
703 800005c2: b9 04 00 1f lgr %r1,%r15
704 800005c6: a7 fb ff 60 aghi %r15,-160
705 800005ca: e3 10 f0 00 00 24 stg %r1,0(%r15)
707 800005d0: a7 29 00 05 lghi %r2,5
708 # brasl allows jumps > 64k & is overkill here bras would do fune
709 800005d4: c0 e5 ff ff ff ee brasl %r14,800005b0 <test>
710 800005da: e3 40 f1 10 00 04 lg %r4,272(%r15)
711 800005e0: eb bf f0 f8 00 04 lmg %r11,%r15,248(%r15)
712 800005e6: 07 f4 br %r4
717 Compiling programs for debugging on Linux for s/390 & z/Architecture
718 ====================================================================
719 -gdwarf-2 now works it should be considered the default debugging
720 format for s/390 & z/Architecture as it is more reliable for debugging
721 shared libraries, normal -g debugging works much better now
722 Thanks to the IBM java compiler developers bug reports.
724 This is typically done adding/appending the flags -g or -gdwarf-2 to the
725 CFLAGS & LDFLAGS variables Makefile of the program concerned.
727 If using gdb & you would like accurate displays of registers &
728 stack traces compile without optimisation i.e make sure
729 that there is no -O2 or similar on the CFLAGS line of the Makefile &
730 the emitted gcc commands, obviously this will produce worse code
731 ( not advisable for shipment ) but it is an aid to the debugging process.
733 This aids debugging because the compiler will copy parameters passed in
734 in registers onto the stack so backtracing & looking at passed in
735 parameters will work, however some larger programs which use inline functions
736 will not compile without optimisation.
738 Debugging with optimisation has since much improved after fixing
739 some bugs, please make sure you are using gdb-5.0 or later developed
749 Addresses & values in the VM debugger are always hex never decimal
750 Address ranges are of the format <HexValue1>-<HexValue2> or
751 <HexValue1>.<HexValue2>
752 For example, the address range 0x2000 to 0x3000 can be described as 2000-3000
755 The VM Debugger is case insensitive.
757 VM's strengths are usually other debuggers weaknesses you can get at any
758 resource no matter how sensitive e.g. memory management resources, change
759 address translation in the PSW. For kernel hacking you will reap dividends if
762 The VM Debugger displays operators but not operands, and also the debugger
763 displays useful information on the same line as the author of the code probably
764 felt that it was a good idea not to go over the 80 columns on the screen.
765 This isn't as unintuitive as it may seem as the s/390 instructions are easy to
766 decode mentally and you can make a good guess at a lot of them as all the
767 operands are nibble (half byte aligned).
768 So if you have an objdump listing by hand, it is quite easy to follow, and if
769 you don't have an objdump listing keep a copy of the s/390 Reference Summary
770 or alternatively the s/390 principles of operation next to you.
771 e.g. even I can guess that
772 0001AFF8' LR 180F CC 0
773 is a ( load register ) lr r0,r15
775 Also it is very easy to tell the length of a 390 instruction from the 2 most
776 significant bits in the instruction (not that this info is really useful except
777 if you are trying to make sense of a hexdump of code).
779 Bits Instruction Length
780 ------------------------------------------
786 The debugger also displays other useful info on the same line such as the
787 addresses being operated on destination addresses of branches & condition codes.
789 00019736' AHI A7DAFF0E CC 1
790 000198BA' BRC A7840004 -> 000198C2' CC 0
791 000198CE' STM 900EF068 >> 0FA95E78 CC 2
795 Useful VM debugger commands
796 ---------------------------
798 I suppose I'd better mention this before I start
799 to list the current active traces do
801 there can be a maximum of 255 of these per set
802 ( more about trace sets later ).
803 To stop traces issue a
805 To delete a particular breakpoint issue
806 TR DEL <breakpoint number>
808 The PA1 key drops to CP mode so you can issue debugger commands,
809 Doing alt c (on my 3270 console at least ) clears the screen.
810 hitting b <enter> comes back to the running operating system
811 from cp mode ( in our case linux ).
812 It is typically useful to add shortcuts to your profile.exec file
813 if you have one ( this is roughly equivalent to autoexec.bat in DOS ).
814 file here are a few from mine.
815 /* this gives me command history on issuing f12 */
819 /* goes to trace set a */
820 set pf1 imm tr goto a
821 /* goes to trace set b */
822 set pf2 imm tr goto b
823 /* goes to trace set c */
824 set pf3 imm tr goto c
830 Setting a simple breakpoint
832 To debug a particular function try
833 TR I R <function address range>
834 TR I on its own will single step.
835 TR I DATA <MNEMONIC> <OPTIONAL RANGE> will trace for particular mnemonics
837 TR I DATA 4D R 0197BC.4000
838 will trace for BAS'es ( opcode 4D ) in the range 0197BC.4000
839 if you were inclined you could add traces for all branch instructions &
840 suffix them with the run prefix so you would have a backtrace on screen
841 when a program crashes.
842 TR BR <INTO OR FROM> will trace branches into or out of an address.
844 TR BR INTO 0 is often quite useful if a program is getting awkward & deciding
845 to branch to 0 & crashing as this will stop at the address before in jumps to 0.
846 TR I R <address range> RUN cmd d g
847 single steps a range of addresses but stays running &
848 displays the gprs on each step.
852 Displaying & modifying Registers
853 --------------------------------
854 D G will display all the gprs
855 Adding a extra G to all the commands is necessary to access the full 64 bit
856 content in VM on z/Architecture. Obviously this isn't required for access
857 registers as these are still 32 bit.
858 e.g. DGG instead of DG
859 D X will display all the control registers
860 D AR will display all the access registers
861 D AR4-7 will display access registers 4 to 7
862 CPU ALL D G will display the GRPS of all CPUS in the configuration
863 D PSW will display the current PSW
864 st PSW 2000 will put the value 2000 into the PSW &
865 cause crash your machine.
866 D PREFIX displays the prefix offset
871 To display memory mapped using the current PSW's mapping try
873 To make VM display a message each time it hits a particular address and
875 D I<range> will disassemble/display a range of instructions.
876 ST addr 32 bit word will store a 32 bit aligned address
877 D T<range> will display the EBCDIC in an address (if you are that way inclined)
878 D R<range> will display real addresses ( without DAT ) but with prefixing.
879 There are other complex options to display if you need to get at say home space
880 but are in primary space the easiest thing to do is to temporarily
881 modify the PSW to the other addressing mode, display the stuff & then
888 If you want to issue a debugger command without halting your virtual machine
889 with the PA1 key try prefixing the command with #CP e.g.
891 also suffixing most debugger commands with RUN will cause them not
892 to stop just display the mnemonic at the current instruction on the console.
893 If you have several breakpoints you want to put into your program &
894 you get fed up of cross referencing with System.map
895 you can do the following trick for several symbols.
896 grep do_signal System.map
897 which emits the following among other things
901 TR I PSWA 0001f4e0 cmd msg * do_signal
902 This sends a message to your own console each time do_signal is entered.
903 ( As an aside I wrote a perl script once which automatically generated a REXX
904 script with breakpoints on every kernel procedure, this isn't a good idea
905 because there are thousands of these routines & VM can only set 255 breakpoints
906 at a time so you nearly had to spend as long pruning the file down as you would
907 entering the msgs by hand), however, the trick might be useful for a single
908 object file. In the 3270 terminal emulator x3270 there is a very useful option
909 in the file menu called "Save Screen In File" - this is very good for keeping a
912 From CMS help <command name> will give you online help on a particular command.
916 Also CP has a file called profile.exec which automatically gets called
917 on startup of CMS ( like autoexec.bat ), keeping on a DOS analogy session
918 CP has a feature similar to doskey, it may be useful for you to
919 use profile.exec to define some keystrokes.
922 This does a single step in VM on pressing F8.
924 This sets up the ^ key.
925 which can be used for ^c (ctrl-c),^z (ctrl-z) which can't be typed directly
926 into some 3270 consoles.
928 This types the starting keystrokes for a sysrq see SysRq below.
930 This retrieves command history on pressing F12.
933 Sometimes in VM the display is set up to scroll automatically this
934 can be very annoying if there are messages you wish to look at
937 This will nearly stop automatic screen updates, however it will
938 cause a denial of service if lots of messages go to the 3270 console,
939 so it would be foolish to use this as the default on a production machine.
942 Tracing particular processes
943 ----------------------------
944 The kernel's text segment is intentionally at an address in memory that it will
945 very seldom collide with text segments of user programs ( thanks Martin ),
946 this simplifies debugging the kernel.
947 However it is quite common for user processes to have addresses which collide
948 this can make debugging a particular process under VM painful under normal
949 circumstances as the process may change when doing a
950 TR I R <address range>.
951 Thankfully after reading VM's online help I figured out how to debug
952 I particular process.
954 Your first problem is to find the STD ( segment table designation )
955 of the program you wish to debug.
956 There are several ways you can do this here are a few
957 1) objdump --syms <program to be debugged> | grep main
958 To get the address of main in the program.
959 tr i pswa <address of main>
960 Start the program, if VM drops to CP on what looks like the entry
961 point of the main function this is most likely the process you wish to debug.
962 Now do a D X13 or D XG13 on z/Architecture.
963 On 31 bit the STD is bits 1-19 ( the STO segment table origin )
964 & 25-31 ( the STL segment table length ) of CR13.
966 TR I R STD <CR13's value> 0.7fffffff
968 TR I R STD 8F32E1FF 0.7fffffff
969 Another very useful variation is
970 TR STORE INTO STD <CR13's value> <address range>
971 for finding out when a particular variable changes.
973 An alternative way of finding the STD of a currently running process
974 is to do the following, ( this method is more complex but
975 could be quite convenient if you aren't updating the kernel much &
976 so your kernel structures will stay constant for a reasonable period of
979 grep task /proc/<pid>/status
980 from this you should see something like
981 task: 0f160000 ksp: 0f161de8 pt_regs: 0f161f68
982 This now gives you a pointer to the task structure.
983 Now make CC:="s390-gcc -g" kernel/sched.s
984 To get the task_struct stabinfo.
985 ( task_struct is defined in include/linux/sched.h ).
986 Now we want to look at
988 on my machine the active_mm in the task structure stab is
989 active_mm:(4,12),672,32
990 its offset is 672/8=84=0x54
991 the pgd member in the mm_struct stab is
992 pgd:(4,6)=*(29,5),96,32
993 so its offset is 96/8=12=0xc
996 hexdump -s 0xf160054 /dev/mem | more
997 i.e. task_struct+active_mm offset
998 to look at the active_mm member
999 f160054 0fee cc60 0019 e334 0000 0000 0000 0011
1000 hexdump -s 0x0feecc6c /dev/mem | more
1001 i.e. active_mm+pgd offset
1002 feecc6c 0f2c 0000 0000 0001 0000 0001 0000 0010
1003 we get something like
1005 TR I R STD <pgd|0x7f> 0.7fffffff
1006 i.e. the 0x7f is added because the pgd only
1007 gives the page table origin & we need to set the low bits
1008 to the maximum possible segment table length.
1009 TR I R STD 0f2c007f 0.7fffffff
1010 on z/Architecture you'll probably need to do
1011 TR I R STD <pgd|0x7> 0.ffffffffffffffff
1012 to set the TableType to 0x1 & the Table length to 3.
1016 Tracing Program Exceptions
1017 --------------------------
1018 If you get a crash which says something like
1019 illegal operation or specification exception followed by a register dump
1020 You can restart linux & trace these using the tr prog <range or value> trace
1024 The most common ones you will normally be tracing for is
1025 1=operation exception
1026 2=privileged operation exception
1027 4=protection exception
1028 5=addressing exception
1029 6=specification exception
1030 10=segment translation exception
1031 11=page translation exception
1033 The full list of these is on page 22 of the current s/390 Reference Summary.
1035 tr prog 10 will trace segment translation exceptions.
1036 tr prog on its own will trace all program interruption codes.
1040 On starting VM you are initially in the INITIAL trace set.
1041 You can do a Q TR to verify this.
1042 If you have a complex tracing situation where you wish to wait for instance
1043 till a driver is open before you start tracing IO, but know in your
1044 heart that you are going to have to make several runs through the code till you
1045 have a clue whats going on.
1048 TR I PSWA <Driver open address>
1049 hit b to continue till breakpoint
1050 reach the breakpoint
1053 TR IO 7c08-7c09 inst int run
1054 or whatever the IO channels you wish to trace are & hit b
1056 To got back to the initial trace set do
1058 & the TR I PSWA <Driver open address> will be the only active breakpoint again.
1061 Tracing linux syscalls under VM
1062 -------------------------------
1063 Syscalls are implemented on Linux for S390 by the Supervisor call instruction
1064 (SVC). There 256 possibilities of these as the instruction is made up of a 0xA
1065 opcode and the second byte being the syscall number. They are traced using the
1067 TR SVC <Optional value or range>
1068 the syscalls are defined in linux/arch/s390/include/asm/unistd.h
1069 e.g. to trace all file opens just do
1070 TR SVC 5 ( as this is the syscall number of open )
1073 SMP Specific commands
1074 ---------------------
1075 To find out how many cpus you have
1076 Q CPUS displays all the CPU's available to your virtual machine
1077 To find the cpu that the current cpu VM debugger commands are being directed at
1078 do Q CPU to change the current cpu VM debugger commands are being directed at do
1079 CPU <desired cpu no>
1081 On a SMP guest issue a command to all CPUs try prefixing the command with cpu
1082 all. To issue a command to a particular cpu try cpu <cpu number> e.g.
1083 CPU 01 TR I R 2000.3000
1084 If you are running on a guest with several cpus & you have a IO related problem
1085 & cannot follow the flow of code but you know it isn't smp related.
1086 from the bash prompt issue
1087 shutdown -h now or halt.
1088 do a Q CPUS to find out how many cpus you have
1089 detach each one of them from cp except cpu 0
1091 DETACH CPU 01-(number of cpus in configuration)
1093 TR SIGP will trace inter processor signal processor instructions.
1094 DEFINE CPU 01-(number in configuration)
1095 will get your guests cpus back.
1098 Help for displaying ascii textstrings
1099 -------------------------------------
1100 On the very latest VM Nucleus'es VM can now display ascii
1101 ( thanks Neale for the hint ) by doing
1108 Under older VM debuggers (I love EBDIC too) you can use following little
1109 program which converts a command line of hex digits to ascii text. It can be
1110 compiled under linux and you can copy the hex digits from your x3270 terminal
1111 to your xterm if you are debugging from a linuxbox.
1113 This is quite useful when looking at a parameter passed in as a text string
1114 under VM ( unless you are good at decoding ASCII in your head ).
1116 e.g. consider tracing an open syscall
1118 We have stopped at a breakpoint
1119 000151B0' SVC 0A05 -> 0001909A' CC 0
1121 D 20.8 to check the SVC old psw in the prefix area and see was it from userspace
1122 (for the layout of the prefix area consult the "Fixed Storage Locations"
1123 chapter of the s/390 Reference Summary if you have it available).
1124 V00000020 070C2000 800151B2
1125 The problem state bit wasn't set & it's also too early in the boot sequence
1126 for it to be a userspace SVC if it was we would have to temporarily switch the
1127 psw to user space addressing so we could get at the first parameter of the open
1132 Now display what gpr2 is pointing to
1134 V00014CB4 2F646576 2F636F6E 736F6C65 00001BF5
1135 V00014CC4 FC00014C B4001001 E0001000 B8070707
1136 Now copy the text till the first 00 hex ( which is the end of the string
1137 to an xterm & do hex2ascii on it.
1138 hex2ascii 2F646576 2F636F6E 736F6C65 00
1140 Decoded Hex:=/ d e v / c o n s o l e 0x00
1141 We were opening the console device,
1143 You can compile the code below yourself for practice :-),
1146 * a useful little tool for converting a hexadecimal command line to ascii
1148 * Author(s): Denis Joseph Barrow (djbarrow@de.ibm.com,barrow_dj@yahoo.com)
1149 * (C) 2000 IBM Deutschland Entwicklung GmbH, IBM Corporation.
1153 int main(int argc,char *argv[])
1155 int cnt1,cnt2,len,toggle=0;
1157 unsigned char c,hex;
1159 if(argc>1&&(strcmp(argv[1],"-a")==0))
1161 printf("Decoded Hex:=");
1162 for(cnt1=startcnt;cnt1<argc;cnt1++)
1164 len=strlen(argv[cnt1]);
1165 for(cnt2=0;cnt2<len;cnt2++)
1185 printf("0x%02X ",(int)hex);
1206 Stack tracing under VM
1207 ----------------------
1211 Here are the tricks I use 9 out of 10 times it works pretty well,
1213 When your backchain reaches a dead end
1214 --------------------------------------
1215 This can happen when an exception happens in the kernel and the kernel is
1216 entered twice. If you reach the NULL pointer at the end of the back chain you
1217 should be able to sniff further back if you follow the following tricks.
1218 1) A kernel address should be easy to recognise since it is in
1219 primary space & the problem state bit isn't set & also
1220 The Hi bit of the address is set.
1221 2) Another backchain should also be easy to recognise since it is an
1222 address pointing to another address approximately 100 bytes or 0x70 hex
1223 behind the current stackpointer.
1226 Here is some practice.
1227 boot the kernel & hit PA1 at some random time
1228 d g to display the gprs, this should display something like
1229 GPR 0 = 00000001 00156018 0014359C 00000000
1230 GPR 4 = 00000001 001B8888 000003E0 00000000
1231 GPR 8 = 00100080 00100084 00000000 000FE000
1232 GPR 12 = 00010400 8001B2DC 8001B36A 000FFED8
1233 Note that GPR14 is a return address but as we are real men we are going to
1235 display 0x40 bytes after the stack pointer.
1237 V000FFED8 000FFF38 8001B838 80014C8E 000FFF38
1238 V000FFEE8 00000000 00000000 000003E0 00000000
1239 V000FFEF8 00100080 00100084 00000000 000FE000
1240 V000FFF08 00010400 8001B2DC 8001B36A 000FFED8
1243 Ah now look at whats in sp+56 (sp+0x38) this is 8001B36A our saved r14 if
1244 you look above at our stackframe & also agrees with GPR14.
1248 we now are taking the contents of SP to get our first backchain.
1250 V000FFF38 000FFFA0 00000000 00014995 00147094
1251 V000FFF48 00147090 001470A0 000003E0 00000000
1252 V000FFF58 00100080 00100084 00000000 001BF1D0
1253 V000FFF68 00010400 800149BA 80014CA6 000FFF38
1255 This displays a 2nd return address of 80014CA6
1257 now do d 000FFFA0.40 for our 3rd backchain
1259 V000FFFA0 04B52002 0001107F 00000000 00000000
1260 V000FFFB0 00000000 00000000 FF000000 0001107F
1261 V000FFFC0 00000000 00000000 00000000 00000000
1262 V000FFFD0 00010400 80010802 8001085A 000FFFA0
1265 our 3rd return address is 8001085A
1267 as the 04B52002 looks suspiciously like rubbish it is fair to assume that the
1268 kernel entry routines for the sake of optimisation don't set up a backchain.
1270 now look at System.map to see if the addresses make any sense.
1272 grep -i 0001b3 System.map
1273 outputs among other things
1276 is cpu_idle+0x66 ( quiet the cpu is asleep, don't wake it )
1279 grep -i 00014 System.map
1280 produces among other things
1281 00014a78 T start_kernel
1282 so 0014CA6 is start_kernel+some hex number I can't add in my head.
1284 grep -i 00108 System.map
1287 so 8001085A is _stext+0x5a
1289 Congrats you've done your first backchain.
1293 s/390 & z/Architecture IO Overview
1294 ==================================
1296 I am not going to give a course in 390 IO architecture as this would take me
1297 quite a while and I'm no expert. Instead I'll give a 390 IO architecture
1298 summary for Dummies. If you have the s/390 principles of operation available
1299 read this instead. If nothing else you may find a few useful keywords in here
1300 and be able to use them on a web search engine to find more useful information.
1302 Unlike other bus architectures modern 390 systems do their IO using mostly
1303 fibre optics and devices such as tapes and disks can be shared between several
1304 mainframes. Also S390 can support up to 65536 devices while a high end PC based
1305 system might be choking with around 64.
1307 Here is some of the common IO terminology:
1310 This is the logical number most IO commands use to talk to an IO device. There
1311 can be up to 0x10000 (65536) of these in a configuration, typically there are a
1312 few hundred. Under VM for simplicity they are allocated contiguously, however
1313 on the native hardware they are not. They typically stay consistent between
1314 boots provided no new hardware is inserted or removed.
1315 Under Linux for s390 we use these as IRQ's and also when issuing an IO command
1316 (CLEAR SUBCHANNEL, HALT SUBCHANNEL, MODIFY SUBCHANNEL, RESUME SUBCHANNEL,
1317 START SUBCHANNEL, STORE SUBCHANNEL and TEST SUBCHANNEL). We use this as the ID
1318 of the device we wish to talk to. The most important of these instructions are
1319 START SUBCHANNEL (to start IO), TEST SUBCHANNEL (to check whether the IO
1320 completed successfully) and HALT SUBCHANNEL (to kill IO). A subchannel can have
1321 up to 8 channel paths to a device, this offers redundancy if one is not
1325 This number remains static and is closely tied to the hardware. There are 65536
1326 of these, made up of a CHPID (Channel Path ID, the most significant 8 bits) and
1327 another lsb 8 bits. These remain static even if more devices are inserted or
1328 removed from the hardware. There is a 1 to 1 mapping between subchannels and
1329 device numbers, provided devices aren't inserted or removed.
1331 Channel Control Words:
1332 CCWs are linked lists of instructions initially pointed to by an operation
1333 request block (ORB), which is initially given to Start Subchannel (SSCH)
1334 command along with the subchannel number for the IO subsystem to process
1335 while the CPU continues executing normal code.
1336 CCWs come in two flavours, Format 0 (24 bit for backward compatibility) and
1337 Format 1 (31 bit). These are typically used to issue read and write (and many
1338 other) instructions. They consist of a length field and an absolute address
1340 Each IO typically gets 1 or 2 interrupts, one for channel end (primary status)
1341 when the channel is idle, and the second for device end (secondary status).
1342 Sometimes you get both concurrently. You check how the IO went on by issuing a
1343 TEST SUBCHANNEL at each interrupt, from which you receive an Interruption
1344 response block (IRB). If you get channel and device end status in the IRB
1345 without channel checks etc. your IO probably went okay. If you didn't you
1346 probably need to examine the IRB, extended status word etc.
1347 If an error occurs, more sophisticated control units have a facility known as
1348 concurrent sense. This means that if an error occurs Extended sense information
1349 will be presented in the Extended status word in the IRB. If not you have to
1350 issue a subsequent SENSE CCW command after the test subchannel.
1353 TPI (Test pending interrupt) can also be used for polled IO, but in
1354 multitasking multiprocessor systems it isn't recommended except for
1355 checking special cases (i.e. non looping checks for pending IO etc.).
1357 Store Subchannel and Modify Subchannel can be used to examine and modify
1358 operating characteristics of a subchannel (e.g. channel paths).
1360 Other IO related Terms:
1361 Sysplex: S390's Clustering Technology
1362 QDIO: S390's new high speed IO architecture to support devices such as gigabit
1363 ethernet, this architecture is also designed to be forward compatible with
1364 upcoming 64 bit machines.
1369 Input Output Processors (IOP's) are responsible for communicating between
1370 the mainframe CPU's & the channel & relieve the mainframe CPU's from the
1371 burden of communicating with IO devices directly, this allows the CPU's to
1372 concentrate on data processing.
1374 IOP's can use one or more links ( known as channel paths ) to talk to each
1375 IO device. It first checks for path availability & chooses an available one,
1376 then starts ( & sometimes terminates IO ).
1377 There are two types of channel path: ESCON & the Parallel IO interface.
1379 IO devices are attached to control units, control units provide the
1380 logic to interface the channel paths & channel path IO protocols to
1381 the IO devices, they can be integrated with the devices or housed separately
1382 & often talk to several similar devices ( typical examples would be raid
1383 controllers or a control unit which connects to 1000 3270 terminals ).
1386 +---------------------------------------------------------------+
1387 | +-----+ +-----+ +-----+ +-----+ +----------+ +----------+ |
1388 | | CPU | | CPU | | CPU | | CPU | | Main | | Expanded | |
1389 | | | | | | | | | | Memory | | Storage | |
1390 | +-----+ +-----+ +-----+ +-----+ +----------+ +----------+ |
1391 |---------------------------------------------------------------+
1393 |---------------------------------------------------------------
1394 | C | C | C | C | C | C | C | C | C | C | C | C | C | C | C | C |
1395 ----------------------------------------------------------------
1397 || Bus & Tag Channel Path || ESCON
1398 || ====================== || Channel
1400 +----------+ +----------+ +----------+
1402 | CU | | CU | | CU |
1404 +----------+ +----------+ +----------+
1406 +----------+ +----------+ +----------+ +----------+ +----------+
1407 |I/O Device| |I/O Device| |I/O Device| |I/O Device| |I/O Device|
1408 +----------+ +----------+ +----------+ +----------+ +----------+
1409 CPU = Central Processing Unit
1414 The 390 IO systems come in 2 flavours the current 390 machines support both
1416 The Older 360 & 370 Interface,sometimes called the Parallel I/O interface,
1417 sometimes called Bus-and Tag & sometimes Original Equipment Manufacturers
1420 This byte wide Parallel channel path/bus has parity & data on the "Bus" cable
1421 and control lines on the "Tag" cable. These can operate in byte multiplex mode
1422 for sharing between several slow devices or burst mode and monopolize the
1423 channel for the whole burst. Up to 256 devices can be addressed on one of these
1424 cables. These cables are about one inch in diameter. The maximum unextended
1425 length supported by these cables is 125 Meters but this can be extended up to
1426 2km with a fibre optic channel extended such as a 3044. The maximum burst speed
1427 supported is 4.5 megabytes per second. However, some really old processors
1428 support only transfer rates of 3.0, 2.0 & 1.0 MB/sec.
1429 One of these paths can be daisy chained to up to 8 control units.
1432 ESCON if fibre optic it is also called FICON
1433 Was introduced by IBM in 1990. Has 2 fibre optic cables and uses either leds or
1434 lasers for communication at a signaling rate of up to 200 megabits/sec. As
1435 10bits are transferred for every 8 bits info this drops to 160 megabits/sec
1436 and to 18.6 Megabytes/sec once control info and CRC are added. ESCON only
1437 operates in burst mode.
1439 ESCONs typical max cable length is 3km for the led version and 20km for the
1440 laser version known as XDF (extended distance facility). This can be further
1441 extended by using an ESCON director which triples the above mentioned ranges.
1442 Unlike Bus & Tag as ESCON is serial it uses a packet switching architecture,
1443 the standard Bus & Tag control protocol is however present within the packets.
1444 Up to 256 devices can be attached to each control unit that uses one of these
1447 Common 390 Devices include:
1448 Network adapters typically OSA2,3172's,2116's & OSA-E gigabit ethernet adapters,
1449 Consoles 3270 & 3215 (a teletype emulated under linux for a line mode console).
1450 DASD's direct access storage devices ( otherwise known as hard disks ).
1452 CTC ( Channel to Channel Adapters ),
1453 ESCON or Parallel Cables used as a very high speed serial link
1457 Debugging IO on s/390 & z/Architecture under VM
1458 ===============================================
1460 Now we are ready to go on with IO tracing commands under VM
1462 A few self explanatory queries:
1465 Q DISK ( This command is CMS specific )
1473 Q OSA on my machine returns
1474 OSA 7C08 ON OSA 7C08 SUBCHANNEL = 0000
1475 OSA 7C09 ON OSA 7C09 SUBCHANNEL = 0001
1476 OSA 7C14 ON OSA 7C14 SUBCHANNEL = 0002
1477 OSA 7C15 ON OSA 7C15 SUBCHANNEL = 0003
1479 If you have a guest with certain privileges you may be able to see devices
1480 which don't belong to you. To avoid this, add the option V.
1484 Now using the device numbers returned by this command we will
1485 Trace the io starting up on the first device 7c08 & 7c09
1486 In our simplest case we can trace the
1488 like TR SSCH 7C08-7C09
1489 or the halt subchannels
1490 or TR HSCH 7C08-7C09
1491 MSCH's ,STSCH's I think you can guess the rest
1493 A good trick is tracing all the IO's and CCWS and spooling them into the reader
1494 of another VM guest so he can ftp the logfile back to his own machine. I'll do
1495 a small bit of this and give you a look at the output.
1497 1) Spool stdout to VM reader
1498 SP PRT TO (another vm guest ) or * for the local vm guest
1499 2) Fill the reader with the trace
1500 TR IO 7c08-7c09 INST INT CCW PRT RUN
1507 6) list reader contents
1509 7) copy it to linux4's minidisk
1510 RECEIVE / LOG TXT A1 ( replace
1512 filel & press F11 to look at it
1513 You should see something like:
1515 00020942' SSCH B2334000 0048813C CC 0 SCH 0000 DEV 7C08
1516 CPA 000FFDF0 PARM 00E2C9C4 KEY 0 FPI C0 LPM 80
1517 CCW 000FFDF0 E4200100 00487FE8 0000 E4240100 ........
1520 00020B0A' I/O DEV 7C08 -> 000197BC' SCH 0000 PARM 00E2C9C4
1521 00021628' TSCH B2354000 >> 00488164 CC 0 SCH 0000 DEV 7C08
1522 CCWA 000FFDF8 DEV STS 0C SCH STS 00 CNT 00EC
1523 KEY 0 FPI C0 CC 0 CTLS 4007
1524 00022238' STSCH B2344000 >> 00488108 CC 0 SCH 0000 DEV 7C08
1526 If you don't like messing up your readed ( because you possibly booted from it )
1527 you can alternatively spool it to another readers guest.
1530 Other common VM device related commands
1531 ---------------------------------------------
1532 These commands are listed only because they have
1533 been of use to me in the past & may be of use to
1534 you too. For more complete info on each of the commands
1535 use type HELP <command> from CMS.
1538 ATT <devno range> <guest>
1539 attach a device to guest * for your own guest
1540 READY <devno> cause VM to issue a fake interrupt.
1542 The VARY command is normally only available to VM administrators.
1543 VARY ON PATH <path> TO <devno range>
1544 VARY OFF PATH <PATH> FROM <devno range>
1545 This is used to switch on or off channel paths to devices.
1547 Q CHPID <channel path ID>
1548 This displays state of devices using this channel path
1549 D SCHIB <subchannel>
1550 This displays the subchannel information SCHIB block for the device.
1551 this I believe is also only available to administrators.
1553 defines a virtual CTC channel to channel connection
1554 2 need to be defined on each guest for the CTC driver to use.
1555 COUPLE devno userid remote devno
1556 Joins a local virtual device to a remote virtual device
1557 ( commonly used for the CTC driver ).
1559 Building a VM ramdisk under CMS which linux can use
1560 def vfb-<blocksize> <subchannel> <number blocks>
1561 blocksize is commonly 4096 for linux.
1563 format <subchannel> <driver letter e.g. x> (blksize <blocksize>
1565 Sharing a disk between multiple guests
1566 LINK userid devno1 devno2 mode password
1572 N.B. if compiling for debugging gdb works better without optimisation
1573 ( see Compiling programs for debugging )
1577 gdb <victim program> <optional corefile>
1581 help: gives help on commands
1585 Note gdb's online help is very good use it.
1590 info registers: displays registers other than floating point.
1591 info all-registers: displays floating points as well.
1592 disassemble: disassembles
1594 disassemble without parameters will disassemble the current function
1595 disassemble $pc $pc+10
1597 Viewing & modifying variables
1598 -----------------------------
1599 print or p: displays variable or register
1600 e.g. p/x $sp will display the stack pointer
1602 display: prints variable or register each time program stops
1604 display/x $pc will display the program counter
1607 undisplay : undo's display's
1609 info breakpoints: shows all current breakpoints
1611 info stack: shows stack back trace (if this doesn't work too well, I'll show
1612 you the stacktrace by hand below).
1614 info locals: displays local variables.
1616 info args: display current procedure arguments.
1618 set args: will set argc & argv each time the victim program is invoked.
1620 set <variable>=value
1628 step: steps n lines of sourcecode
1630 step 100 steps 100 lines of code.
1632 next: like step except this will not step into subroutines
1634 stepi: steps a single machine code instruction.
1637 nexti: steps a single machine code instruction but will not step into
1640 finish: will run until exit of the current routine
1642 run: (re)starts a program
1644 cont: continues a program
1662 Here's a really useful one for large programs
1664 Set a breakpoint for all functions matching REGEXP
1667 will set a breakpoint with all functions with 390 in their name.
1670 lists all breakpoints
1672 delete: delete breakpoint by number or delete them all
1674 delete 1 will delete the first breakpoint
1675 delete will delete them all
1677 watch: This will set a watchpoint ( usually hardware assisted ),
1678 This will watch a variable till it changes
1680 watch cnt, will watch the variable cnt till it changes.
1681 As an aside unfortunately gdb's, architecture independent watchpoint code
1682 is inconsistent & not very good, watchpoints usually work but not always.
1684 info watchpoints: Display currently active watchpoints
1686 condition: ( another useful one )
1687 Specify breakpoint number N to break only if COND is true.
1688 Usage is `condition N COND', where N is an integer and COND is an
1689 expression to be evaluated whenever breakpoint N is reached.
1693 User defined functions/macros
1694 -----------------------------
1695 define: ( Note this is very very useful,simple & powerful )
1696 usage define <name> <list of commands> end
1698 examples which you should consider putting into .gdbinit in your home directory
1701 disassemble $pc $pc+10
1706 disassemble $pc $pc+10
1710 Other hard to classify stuff
1711 ----------------------------
1713 sends the victim program a signal.
1714 e.g. signal 3 will send a SIGQUIT.
1717 what gdb does when the victim receives certain signals.
1721 list lists current function source
1722 list 1,10 list first 10 lines of current file.
1727 Adds directories to be searched for source if gdb cannot find the source.
1728 (note it is a bit sensitive about slashes)
1729 e.g. To add the root of the filesystem to the searchpath do
1734 This calls a function in the victim program, this is pretty powerful
1736 (gdb) call printf("hello world")
1740 You might now be thinking that the line above didn't work, something extra had
1742 (gdb) call fflush(stdout)
1744 As an aside the debugger also calls malloc & free under the hood
1745 to make space for the "hello world" string.
1751 1) command completion works just like bash
1752 ( if you are a bad typist like me this really helps )
1753 e.g. hit br <TAB> & cursor up & down :-).
1755 2) if you have a debugging problem that takes a few steps to recreate
1756 put the steps into a file called .gdbinit in your current working directory
1757 if you have defined a few extra useful user defined commands put these in
1758 your home directory & they will be read each time gdb is launched.
1760 A typical .gdbinit file might be.
1763 break runtime_exception
1767 stack chaining in gdb by hand
1768 -----------------------------
1769 This is done using a the same trick described for VM
1770 p/x (*($sp+56))&0x7fffffff get the first backchain.
1773 Replace 56 with 112 & ignore the &0x7fffffff
1774 in the macros below & do nasty casts to longs like the following
1775 as gdb unfortunately deals with printed arguments as ints which
1776 messes up everything.
1777 i.e. here is a 3rd backchain dereference
1778 p/x *(long *)(***(long ***)$sp+112)
1785 info symbol (*($sp+56))&0x7fffffff
1786 you might see something like.
1787 rl_getc + 36 in section .text telling you what is located at address 0x528f18
1789 p/x (*(*$sp+56))&0x7fffffff
1793 info symbol (*(*$sp+56))&0x7fffffff
1794 rl_read_key + 180 in section .text
1796 p/x (*(**$sp+56))&0x7fffffff
1799 Disassembling instructions without debug info
1800 ---------------------------------------------
1801 gdb typically complains if there is a lack of debugging
1802 symbols in the disassemble command with
1803 "No function contains specified address." To get around
1805 x/<number lines to disassemble>xi <address>
1811 Note: Remember gdb has history just like bash you don't need to retype the
1812 whole line just use the up & down arrows.
1818 From your linuxbox do
1819 man gdb or info gdb.
1824 A core dump is a file generated by the kernel (if allowed) which contains the
1825 registers and all active pages of the program which has crashed.
1826 From this file gdb will allow you to look at the registers, stack trace and
1827 memory of the program as if it just crashed on your system. It is usually
1828 called core and created in the current working directory.
1829 This is very useful in that a customer can mail a core dump to a technical
1830 support department and the technical support department can reconstruct what
1831 happened. Provided they have an identical copy of this program with debugging
1832 symbols compiled in and the source base of this build is available.
1833 In short it is far more useful than something like a crash log could ever hope
1836 Why have I never seen one ?.
1837 Probably because you haven't used the command
1838 ulimit -c unlimited in bash
1839 to allow core dumps, now do
1841 to verify that the limit was accepted.
1844 To create this I'm going to do
1847 to launch gdb (my victim app. ) now be bad & do the following from another
1848 telnet/xterm session to the same machine
1850 kill -SIGSEGV <gdb's pid>
1851 or alternatively use killall -SIGSEGV gdb if you have the killall command.
1852 Now look at the core dump.
1854 Displays the following
1856 Copyright 1998 Free Software Foundation, Inc.
1857 GDB is free software, covered by the GNU General Public License, and you are
1858 welcome to change it and/or distribute copies of it under certain conditions.
1859 Type "show copying" to see the conditions.
1860 There is absolutely no warranty for GDB. Type "show warranty" for details.
1861 This GDB was configured as "s390-ibm-linux"...
1862 Core was generated by `./gdb'.
1863 Program terminated with signal 11, Segmentation fault.
1864 Reading symbols from /usr/lib/libncurses.so.4...done.
1865 Reading symbols from /lib/libm.so.6...done.
1866 Reading symbols from /lib/libc.so.6...done.
1867 Reading symbols from /lib/ld-linux.so.2...done.
1868 #0 0x40126d1a in read () from /lib/libc.so.6
1869 Setting up the environment for debugging gdb.
1870 Breakpoint 1 at 0x4dc6f8: file utils.c, line 471.
1871 Breakpoint 2 at 0x4d87a4: file top.c, line 2609.
1872 (top-gdb) info stack
1873 #0 0x40126d1a in read () from /lib/libc.so.6
1874 #1 0x528f26 in rl_getc (stream=0x7ffffde8) at input.c:402
1875 #2 0x528ed0 in rl_read_key () at input.c:381
1876 #3 0x5167e6 in readline_internal_char () at readline.c:454
1877 #4 0x5168ee in readline_internal_charloop () at readline.c:507
1878 #5 0x51692c in readline_internal () at readline.c:521
1879 #6 0x5164fe in readline (prompt=0x7ffff810)
1881 #7 0x4d7a8a in command_line_input (prompt=0x564420 "(gdb) ", repeat=1,
1882 annotation_suffix=0x4d6b44 "prompt") at top.c:2091
1883 #8 0x4d6cf0 in command_loop () at top.c:1345
1884 #9 0x4e25bc in main (argc=1, argv=0x7ffffdf4) at main.c:635
1889 This is a program which lists the shared libraries which a library needs,
1890 Note you also get the relocations of the shared library text segments which
1891 help when using objdump --source.
1895 libncurses.so.4 => /usr/lib/libncurses.so.4 (0x40018000)
1896 libm.so.6 => /lib/libm.so.6 (0x4005e000)
1897 libc.so.6 => /lib/libc.so.6 (0x40084000)
1898 /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)
1901 Debugging shared libraries
1902 ==========================
1903 Most programs use shared libraries, however it can be very painful
1904 when you single step instruction into a function like printf for the
1905 first time & you end up in functions like _dl_runtime_resolve this is
1906 the ld.so doing lazy binding, lazy binding is a concept in ELF where
1907 shared library functions are not loaded into memory unless they are
1908 actually used, great for saving memory but a pain to debug.
1909 To get around this either relink the program -static or exit gdb type
1910 export LD_BIND_NOW=true this will stop lazy binding & restart the gdb'ing
1911 the program in question.
1917 As modules are dynamically loaded into the kernel their address can be
1918 anywhere to get around this use the -m option with insmod to emit a load
1919 map which can be piped into a file if required.
1921 The proc file system
1922 ====================
1924 It is a filesystem created by the kernel with files which are created on demand
1925 by the kernel if read, or can be used to modify kernel parameters,
1926 it is a powerful concept.
1930 cat /proc/sys/net/ipv4/ip_forward
1931 On my machine outputs
1933 telling me ip_forwarding is not on to switch it on I can do
1934 echo 1 > /proc/sys/net/ipv4/ip_forward
1936 cat /proc/sys/net/ipv4/ip_forward
1937 On my machine now outputs
1939 IP forwarding is on.
1940 There is a lot of useful info in here best found by going in and having a look
1941 around, so I'll take you through some entries I consider important.
1943 All the processes running on the machine have their own entry defined by
1945 So lets have a look at the init process
1953 This contains numerical entries of all the open files,
1954 some of these you can cat e.g. stdout (2)
1959 00400000-00478000 r-xp 00000000 5f:00 4103 /bin/bash
1960 00478000-0047e000 rw-p 00077000 5f:00 4103 /bin/bash
1961 0047e000-00492000 rwxp 00000000 00:00 0
1962 40000000-40015000 r-xp 00000000 5f:00 14382 /lib/ld-2.1.2.so
1963 40015000-40016000 rw-p 00014000 5f:00 14382 /lib/ld-2.1.2.so
1964 40016000-40017000 rwxp 00000000 00:00 0
1965 40017000-40018000 rw-p 00000000 00:00 0
1966 40018000-4001b000 r-xp 00000000 5f:00 14435 /lib/libtermcap.so.2.0.8
1967 4001b000-4001c000 rw-p 00002000 5f:00 14435 /lib/libtermcap.so.2.0.8
1968 4001c000-4010d000 r-xp 00000000 5f:00 14387 /lib/libc-2.1.2.so
1969 4010d000-40111000 rw-p 000f0000 5f:00 14387 /lib/libc-2.1.2.so
1970 40111000-40114000 rw-p 00000000 00:00 0
1971 40114000-4011e000 r-xp 00000000 5f:00 14408 /lib/libnss_files-2.1.2.so
1972 4011e000-4011f000 rw-p 00009000 5f:00 14408 /lib/libnss_files-2.1.2.so
1973 7fffd000-80000000 rwxp ffffe000 00:00 0
1976 Showing us the shared libraries init uses where they are in memory
1977 & memory access permissions for each virtual memory area.
1979 /proc/1/cwd is a softlink to the current working directory.
1980 /proc/1/root is the root of the filesystem for this process.
1982 /proc/1/mem is the current running processes memory which you
1983 can read & write to like a file.
1984 strace uses this sometimes as it is a bit faster than the
1985 rather inefficient ptrace interface for peeking at DATA.
2004 SigPnd: 0000000000000000
2005 SigBlk: 0000000000000000
2006 SigIgn: 7fffffffd7f0d8fc
2007 SigCgt: 00000000280b2603
2008 CapInh: 00000000fffffeff
2009 CapPrm: 00000000ffffffff
2010 CapEff: 00000000fffffeff
2012 User PSW: 070de000 80414146
2013 task: 004b6000 tss: 004b62d8 ksp: 004b7ca8 pt_regs: 004b7f68
2015 00000400 00000000 0000000b 7ffffa90
2016 00000000 00000000 00000000 0045d9f4
2017 0045cafc 7ffffa90 7fffff18 0045cb08
2018 00010400 804039e8 80403af8 7ffff8b0
2020 00000000 00000000 00000000 00000000
2021 00000001 00000000 00000000 00000000
2022 00000000 00000000 00000000 00000000
2023 00000000 00000000 00000000 00000000
2024 Kernel BackChain CallChain BackChain CallChain
2025 004b7ca8 8002bd0c 004b7d18 8002b92c
2026 004b7db8 8005cd50 004b7e38 8005d12a
2028 Showing among other things memory usage & status of some signals &
2029 the processes'es registers from the kernel task_structure
2030 as well as a backchain which may be useful if a process crashes
2031 in the kernel for some unknown reason.
2033 Some driver debugging techniques
2034 ================================
2037 Some of our drivers now support a "debug feature" in
2038 /proc/s390dbf see s390dbf.txt in the linux/Documentation directory
2041 to switch on the lcs "debug feature"
2042 echo 5 > /proc/s390dbf/lcs/level
2043 & then after the error occurred.
2044 cat /proc/s390dbf/lcs/sprintf >/logfile
2045 the logfile now contains some information which may help
2046 tech support resolve a problem in the field.
2050 high level debugging network drivers
2051 ------------------------------------
2052 ifconfig is a quite useful command
2053 it gives the current state of network drivers.
2055 If you suspect your network device driver is dead
2056 one way to check is type
2057 ifconfig <network device>
2059 You should see something like
2060 tr0 Link encap:16/4 Mbps Token Ring (New) HWaddr 00:04:AC:20:8E:48
2061 inet addr:9.164.185.132 Bcast:9.164.191.255 Mask:255.255.224.0
2062 UP BROADCAST RUNNING MULTICAST MTU:2000 Metric:1
2063 RX packets:246134 errors:0 dropped:0 overruns:0 frame:0
2064 TX packets:5 errors:0 dropped:0 overruns:0 carrier:0
2065 collisions:0 txqueuelen:100
2067 if the device doesn't say up
2069 /etc/rc.d/init.d/network start
2070 ( this starts the network stack & hopefully calls ifconfig tr0 up ).
2071 ifconfig looks at the output of /proc/net/dev and presents it in a more
2073 Now ping the device from a machine in the same subnet.
2074 if the RX packets count & TX packets counts don't increment you probably
2078 Do you see any hardware addresses in the cache if not you may have problems.
2080 ping -c 5 <broadcast_addr> i.e. the Bcast field above in the output of
2081 ifconfig. Do you see any replies from machines other than the local machine
2082 if not you may have problems. also if the TX packets count in ifconfig
2083 hasn't incremented either you have serious problems in your driver
2084 (e.g. the txbusy field of the network device being stuck on )
2085 or you may have multiple network devices connected.
2090 There is a new device layer for channel devices, some
2091 drivers e.g. lcs are registered with this layer.
2092 If the device uses the channel device layer you'll be
2093 able to find what interrupts it uses & the current state
2095 See the manpage chandev.8 &type cat /proc/chandev for more info.
2100 This is now supported by linux for s/390 & z/Architecture.
2101 To enable it do compile the kernel with
2102 Kernel Hacking -> Magic SysRq Key Enabled
2103 echo "1" > /proc/sys/kernel/sysrq
2105 echo "8" >/proc/sys/kernel/printk
2106 To make printk output go to console.
2107 On 390 all commands are prefixed with
2110 ^-t will show tasks.
2111 ^-? or some unknown command will display help.
2112 The sysrq key reading is very picky ( I have to type the keys in an
2113 xterm session & paste them into the x3270 console )
2114 & it may be wise to predefine the keys as described in the VM hints above
2116 This is particularly useful for syncing disks unmounting & rebooting
2117 if the machine gets partially hung.
2119 Read Documentation/admin-guide/sysrq.rst for more info
2123 Enterprise Systems Architecture Reference Summary
2124 Enterprise Systems Architecture Principles of Operation
2125 Hartmut Penners s390 stack frame sheet.
2126 IBM Mainframe Channel Attachment a technology brief from a CISCO webpage
2127 Various bits of man & info pages of Linux.
2129 Various info & man pages.
2130 CMS Help on tracing commands.
2131 Linux for s/390 Elf Application Binary Interface
2132 Linux for z/Series Elf Application Binary Interface ( Both Highly Recommended )
2133 z/Architecture Principles of Operation SA22-7832-00
2134 Enterprise Systems Architecture/390 Reference Summary SA22-7209-01 & the
2135 Enterprise Systems Architecture/390 Principles of Operation SA22-7201-05
2139 Special thanks to Neale Ferguson who maintains a much
2140 prettier HTML version of this page at
2141 http://linuxvm.org/penguinvm/
2142 Bob Grainger Stefan Bader & others for reporting bugs
projects / releases.git / blob