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
11 Linux for s/390 & z/Architecture. It isn't intended as a complete reference & 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
29 Figuring out gcc compile errors
35 s/390 & z/Architecture IO Overview
36 Debugging IO on s/390 & z/Architecture under VM
37 GDB on s/390 & z/Architecture
38 Stack chaining in gdb by hand
43 Starting points for debugging scripting languages etc.
50 The current architectures have the following registers.
52 16 General propose registers, 32 bit on s/390 64 bit on z/Architecture, r0-r15 or gpr0-gpr15 used for arithmetic & addressing.
54 16 Control registers, 32 bit on s/390 64 bit on z/Architecture, ( cr0-cr15 kernel usage only ) used for memory management,
55 interrupt control,debugging control etc.
57 16 Access registers ( ar0-ar15 ) 32 bit on s/390 & z/Architecture
58 not used by normal programs but potentially could
59 be used as temporary storage. Their main purpose is their 1 to 1
60 association with general purpose registers and are used in
61 the kernel for copying data between kernel & user address spaces.
62 Access register 0 ( & access register 1 on z/Architecture ( needs 64 bit
63 pointer ) ) is currently used by the pthread library as a pointer to
64 the current running threads private area.
66 16 64 bit floating point registers (fp0-fp15 ) IEEE & HFP floating
67 point format compliant on G5 upwards & a Floating point control reg (FPC)
68 4 64 bit registers (fp0,fp2,fp4 & fp6) HFP only on older machines.
70 Linux (currently) always uses IEEE & emulates G5 IEEE format on older machines,
71 ( provided the kernel is configured for this ).
74 The PSW is the most important register on the machine it
75 is 64 bit on s/390 & 128 bit on z/Architecture & serves the roles of
76 a program counter (pc), condition code register,memory space designator.
77 In IBM standard notation I am counting bit 0 as the MSB.
78 It has several advantages over a normal program counter
79 in that you can change address translation & program counter
80 in a single instruction. To change address translation,
81 e.g. switching address translation off requires that you
82 have a logical=physical mapping for the address you are
87 0 0 Reserved ( must be 0 ) otherwise specification exception occurs.
89 1 1 Program Event Recording 1 PER enabled,
90 PER is used to facilitate debugging e.g. single stepping.
92 2-4 2-4 Reserved ( must be 0 ).
94 5 5 Dynamic address translation 1=DAT on.
96 6 6 Input/Output interrupt Mask
98 7 7 External interrupt Mask used primarily for interprocessor signalling &
101 8-11 8-11 PSW Key used for complex memory protection mechanism not used under linux
103 12 12 1 on s/390 0 on z/Architecture
105 13 13 Machine Check Mask 1=enable machine check interrupts
107 14 14 Wait State set this to 1 to stop the processor except for interrupts & give
108 time to other LPARS used in CPU idle in the kernel to increase overall
109 usage of processor resources.
111 15 15 Problem state ( if set to 1 certain instructions are disabled )
112 all linux user programs run with this bit 1
113 ( useful info for debugging under VM ).
115 16-17 16-17 Address Space Control
117 00 Primary Space Mode when DAT on
118 The linux kernel currently runs in this mode, CR1 is affiliated with
119 this mode & points to the primary segment table origin etc.
121 01 Access register mode this mode is used in functions to
122 copy data between kernel & user space.
124 10 Secondary space mode not used in linux however CR7 the
125 register affiliated with this mode is & this & normally
126 CR13=CR7 to allow us to copy data between kernel & user space.
127 We do this as follows:
128 We set ar2 to 0 to designate its
129 affiliated gpr ( gpr2 )to point to primary=kernel space.
130 We set ar4 to 1 to designate its
131 affiliated gpr ( gpr4 ) to point to secondary=home=user space
132 & then essentially do a memcopy(gpr2,gpr4,size) to
133 copy data between the address spaces, the reason we use home space for the
134 kernel & don't keep secondary space free is that code will not run in
137 11 Home Space Mode all user programs run in this mode.
138 it is affiliated with CR13.
140 18-19 18-19 Condition codes (CC)
142 20 20 Fixed point overflow mask if 1=FPU exceptions for this event
145 21 21 Decimal overflow mask if 1=FPU exceptions for this event occur
148 22 22 Exponent underflow mask if 1=FPU exceptions for this event occur
151 23 23 Significance Mask if 1=FPU exceptions for this event occur
154 24-31 24-30 Reserved Must be 0.
156 31 Extended Addressing Mode
157 32 Basic Addressing Mode
158 Used to set addressing mode
164 32 1=31 bit addressing mode 0=24 bit addressing mode (for backward
165 compatibility), linux always runs with this bit set to 1
167 33-64 Instruction address.
168 33-63 Reserved must be 0
170 In 24 bits mode bits 64-103=0 bits 104-127 Address
171 In 31 bits mode bits 64-96=0 bits 97-127 Address
172 Note: unlike 31 bit mode on s/390 bit 96 must be zero
173 when loading the address with LPSWE otherwise a
174 specification exception occurs, LPSW is fully backward
180 This per cpu memory area is too intimately tied to the processor not to mention.
181 It exists between the real addresses 0-4096 on s/390 & 0-8192 z/Architecture & is exchanged
182 with a 1 page on s/390 or 2 pages on z/Architecture in absolute storage by the set
183 prefix instruction in linux'es startup.
184 This page is mapped to a different prefix for each processor in an SMP configuration
185 ( assuming the os designer is sane of course :-) ).
186 Bytes 0-512 ( 200 hex ) on s/390 & 0-512,4096-4544,4604-5119 currently on z/Architecture
187 are used by the processor itself for holding such information as exception indications &
188 entry points for exceptions.
189 Bytes after 0xc00 hex are used by linux for per processor globals on s/390 & z/Architecture
190 ( there is a gap on z/Architecture too currently between 0xc00 & 1000 which linux uses ).
191 The closest thing to this on traditional architectures is the interrupt
192 vector table. This is a good thing & does simplify some of the kernel coding
193 however it means that we now cannot catch stray NULL pointers in the
194 kernel without hard coded checks.
198 Address Spaces on Intel Linux
199 =============================
201 The traditional Intel Linux is approximately mapped as follows forgive
203 0xFFFFFFFF 4GB Himem *****************
207 ***************** ****************
208 User Space Himem (typically 0xC0000000 3GB )* User Stack * * *
209 ***************** * *
210 * Shared Libs * * Next Process *
211 ***************** * to *
217 0x00000000 ***************** ****************
219 Now it is easy to see that on Intel it is quite easy to recognise a kernel address
220 as being one greater than user space himem ( in this case 0xC0000000).
221 & addresses of less than this are the ones in the current running program on this
222 processor ( if an smp box ).
223 If using the virtual machine ( VM ) as a debugger it is quite difficult to
224 know which user process is running as the address space you are looking at
225 could be from any process in the run queue.
227 The limitation of Intels addressing technique is that the linux
228 kernel uses a very simple real address to virtual addressing technique
229 of Real Address=Virtual Address-User Space Himem.
230 This means that on Intel the kernel linux can typically only address
231 Himem=0xFFFFFFFF-0xC0000000=1GB & this is all the RAM these machines
233 They can lower User Himem to 2GB or lower & thus be
234 able to use 2GB of RAM however this shrinks the maximum size
235 of User Space from 3GB to 2GB they have a no win limit of 4GB unless
239 On 390 our limitations & strengths make us slightly different.
240 For backward compatibility we are only allowed use 31 bits (2GB)
241 of our 32 bit addresses, however, we use entirely separate address
242 spaces for the user & kernel.
244 This means we can support 2GB of non Extended RAM on s/390, & more
245 with the Extended memory management swap device &
246 currently 4TB of physical memory currently on z/Architecture.
249 Address Spaces on Linux for s/390 & z/Architecture
250 ==================================================
252 Our addressing scheme is as follows
255 Himem 0x7fffffff 2GB on s/390 ***************** ****************
256 currently 0x3ffffffffff (2^42)-1 * User Stack * * *
257 on z/Architecture. ***************** * *
259 ***************** * *
265 0x00000000 ***************** ****************
267 This also means that we need to look at the PSW problem state bit
268 or the addressing mode to decide whether we are looking at
269 user or kernel space.
271 Virtual Addresses on s/390 & z/Architecture
272 ===========================================
274 A virtual address on s/390 is made up of 3 parts
275 The SX ( segment index, roughly corresponding to the PGD & PMD in linux terminology )
277 The PX ( page index, corresponding to the page table entry (pte) in linux terminology )
279 The remaining bits BX (the byte index are the offset in the page )
282 On z/Architecture in linux we currently make up an address from 4 parts.
283 The region index bits (RX) 0-32 we currently use bits 22-32
284 The segment index (SX) being bits 33-43
285 The page index (PX) being bits 44-51
286 The byte index (BX) being bits 52-63
289 1) s/390 has no PMD so the PMD is really the PGD also.
290 A lot of this stuff is defined in pgtable.h.
292 2) Also seeing as s/390's page indexes are only 1k in size
293 (bits 12-19 x 4 bytes per pte ) we use 1 ( page 4k )
294 to make the best use of memory by updating 4 segment indices
295 entries each time we mess with a PMD & use offsets
296 0,1024,2048 & 3072 in this page as for our segment indexes.
297 On z/Architecture our page indexes are now 2k in size
298 ( bits 12-19 x 8 bytes per pte ) we do a similar trick
299 but only mess with 2 segment indices each time we mess with
302 3) As z/Architecture supports up to a massive 5-level page table lookup we
303 can only use 3 currently on Linux ( as this is all the generic kernel
304 currently supports ) however this may change in future
305 this allows us to access ( according to my sums )
306 4TB of virtual storage per process i.e.
307 4096*512(PTES)*1024(PMDS)*2048(PGD) = 4398046511104 bytes,
308 enough for another 2 or 3 of years I think :-).
309 to do this we use a region-third-table designation type in
310 our address space control registers.
313 The Linux for s/390 & z/Architecture Kernel Task Structure
314 ==========================================================
315 Each process/thread under Linux for S390 has its own kernel task_struct
316 defined in linux/include/linux/sched.h
317 The S390 on initialisation & resuming of a process on a cpu sets
318 the __LC_KERNEL_STACK variable in the spare prefix area for this cpu
319 (which we use for per-processor globals).
321 The kernel stack pointer is intimately tied with the task structure for
322 each processor as follows.
325 ************************
326 * 1 page kernel stack *
328 ************************
329 * 1 page task_struct *
331 8K aligned ************************
334 ************************
335 * 2 page kernel stack *
337 ************************
338 * 2 page task_struct *
340 16K aligned ************************
342 What this means is that we don't need to dedicate any register or global variable
343 to point to the current running process & can retrieve it with the following
344 very simple construct for s/390 & one very similar for z/Architecture.
346 static inline struct task_struct * get_current(void)
348 struct task_struct *current;
349 __asm__("lhi %0,-8192\n\t"
355 i.e. just anding the current kernel stack pointer with the mask -8192.
356 Thankfully because Linux doesn't have support for nested IO interrupts
357 & our devices have large buffers can survive interrupts being shut for
358 short amounts of time we don't need a separate stack for interrupts.
363 Register Usage & Stackframes on Linux for s/390 & z/Architecture
364 =================================================================
367 This is the code that gcc produces at the top & the bottom of
368 each function. It usually is fairly consistent & similar from
369 function to function & if you know its layout you can probably
370 make some headway in finding the ultimate cause of a problem
371 after a crash without a source level debugger.
373 Note: To follow stackframes requires a knowledge of C or Pascal &
374 limited knowledge of one assembly language.
376 It should be noted that there are some differences between the
377 s/390 & z/Architecture stack layouts as the z/Architecture stack layout didn't have
378 to maintain compatibility with older linkage formats.
383 This is a built in compiler function for runtime allocation
384 of extra space on the callers stack which is obviously freed
385 up on function exit ( e.g. the caller may choose to allocate nothing
386 of a buffer of 4k if required for temporary purposes ), it generates
387 very efficient code ( a few cycles ) when compared to alternatives
390 automatics: These are local variables on the stack,
391 i.e they aren't in registers & they aren't static.
394 This is a pointer to the stack pointer before entering a
395 framed functions ( see frameless function ) prologue got by
396 dereferencing the address of the current stack pointer,
397 i.e. got by accessing the 32 bit value at the stack pointers
401 This is a pointer to the back of the literal pool which
402 is an area just behind each procedure used to store constants
405 call-clobbered: The caller probably needs to save these registers if there
406 is something of value in them, on the stack or elsewhere before making a
407 call to another procedure so that it can restore it later.
410 The code generated by the compiler to return to the caller.
413 A frameless function in Linux for s390 & z/Architecture is one which doesn't
414 need more than the register save area ( 96 bytes on s/390, 160 on z/Architecture )
415 given to it by the caller.
416 A frameless function never:
417 1) Sets up a back chain.
419 3) Calls other normal functions
423 This is a pointer to the global-offset-table in ELF
424 ( Executable Linkable Format, Linux'es most common executable format ),
425 all globals & shared library objects are found using this pointer.
428 ELF shared libraries are typically only loaded when routines in the shared
429 library are actually first called at runtime. This is lazy binding.
431 procedure-linkage-table
432 This is a table found from the GOT which contains pointers to routines
433 in other shared libraries which can't be called to by easier means.
436 The code generated by the compiler to set up the stack frame.
439 This is extra area allocated on the stack of the calling function if the
440 parameters for the callee's cannot all be put in registers, the same
441 area can be reused by each function the caller calls.
444 A COFF executable format based concept of a procedure reference
445 actually being 8 bytes or more as opposed to a simple pointer to the routine.
446 This is typically defined as follows
447 Routine Descriptor offset 0=Pointer to Function
448 Routine Descriptor offset 4=Pointer to Table of Contents
449 The table of contents/TOC is roughly equivalent to a GOT pointer.
450 & it means that shared libraries etc. can be shared between several
451 environments each with their own TOC.
454 static-chain: This is used in nested functions a concept adopted from pascal
455 by gcc not used in ansi C or C++ ( although quite useful ), basically it
456 is a pointer used to reference local variables of enclosing functions.
457 You might come across this stuff once or twice in your lifetime.
460 The function below should return 11 though gcc may get upset & toss warnings
461 about unused variables.
474 s/390 & z/Architecture Register usage
475 =====================================
476 r0 used by syscalls/assembly call-clobbered
477 r1 used by syscalls/assembly call-clobbered
478 r2 argument 0 / return value 0 call-clobbered
479 r3 argument 1 / return value 1 (if long long) call-clobbered
480 r4 argument 2 call-clobbered
481 r5 argument 3 call-clobbered
483 r7 pointer-to arguments 5 to ... saved
486 r10 static-chain ( if nested function ) saved
487 r11 frame-pointer ( if function used alloca ) saved
488 r12 got-pointer saved
489 r13 base-pointer saved
490 r14 return-address saved
491 r15 stack-pointer saved
493 f0 argument 0 / return value ( float/double ) call-clobbered
494 f2 argument 1 call-clobbered
495 f4 z/Architecture argument 2 saved
496 f6 z/Architecture argument 3 saved
497 The remaining floating points
498 f1,f3,f5 f7-f15 are call-clobbered.
502 1) The only requirement is that registers which are used
503 by the callee are saved, e.g. the compiler is perfectly
504 capable of using r11 for purposes other than a frame a
505 frame pointer if a frame pointer is not needed.
506 2) In functions with variable arguments e.g. printf the calling procedure
507 is identical to one without variable arguments & the same number of
508 parameters. However, the prologue of this function is somewhat more
509 hairy owing to it having to move these parameters to the stack to
510 get va_start, va_arg & va_end to work.
511 3) Access registers are currently unused by gcc but are used in
512 the kernel. Possibilities exist to use them at the moment for
513 temporary storage but it isn't recommended.
514 4) Only 4 of the floating point registers are used for
515 parameter passing as older machines such as G3 only have only 4
516 & it keeps the stack frame compatible with other compilers.
517 However with IEEE floating point emulation under linux on the
518 older machines you are free to use the other 12.
519 5) A long long or double parameter cannot be have the
520 first 4 bytes in a register & the second four bytes in the
521 outgoing args area. It must be purely in the outgoing args
522 area if crossing this boundary.
523 6) Floating point parameters are mixed with outgoing args
524 on the outgoing args area in the order the are passed in as parameters.
525 7) Floating point arguments 2 & 3 are saved in the outgoing args area for
532 0 0 back chain ( a 0 here signifies end of back chain )
533 4 8 eos ( end of stack, not used on Linux for S390 used in other linkage formats )
534 8 16 glue used in other s/390 linkage formats for saved routine descriptors etc.
535 12 24 glue used in other s/390 linkage formats for saved routine descriptors etc.
538 24 48 saved r6 of caller function
539 28 56 saved r7 of caller function
540 32 64 saved r8 of caller function
541 36 72 saved r9 of caller function
542 40 80 saved r10 of caller function
543 44 88 saved r11 of caller function
544 48 96 saved r12 of caller function
545 52 104 saved r13 of caller function
546 56 112 saved r14 of caller function
547 60 120 saved r15 of caller function
548 64 128 saved f4 of caller function
549 72 132 saved f6 of caller function
551 96 160 outgoing args passed from caller to callee
552 96+x 160+x possible stack alignment ( 8 bytes desirable )
553 96+x+y 160+x+y alloca space of caller ( if used )
554 96+x+y+z 160+x+y+z automatics of caller ( if used )
557 A sample program with comments.
558 ===============================
560 Comments on the function test
561 -----------------------------
562 1) It didn't need to set up a pointer to the constant pool gpr13 as it isn't used
564 2) This is a frameless function & no stack is bought.
565 3) The compiler was clever enough to recognise that it could return the
566 value in r2 as well as use it for the passed in parameter ( :-) ).
567 4) The basr ( branch relative & save ) trick works as follows the instruction
568 has a special case with r0,r0 with some instruction operands is understood as
569 the literal value 0, some risc architectures also do this ). So now
570 we are branching to the next address & the address new program counter is
571 in r13,so now we subtract the size of the function prologue we have executed
572 + the size of the literal pool to get to the top of the literal pool
573 0040037c int test(int b)
574 { # Function prologue below
575 40037c: 90 de f0 34 stm %r13,%r14,52(%r15) # Save registers r13 & r14
576 400380: 0d d0 basr %r13,%r0 # Set up pointer to constant pool using
577 400382: a7 da ff fa ahi %r13,-6 # basr trick
580 400386: a7 2a 00 05 ahi %r2,5 # add 5 to r2
582 # Function epilogue below
583 40038a: 98 de f0 34 lm %r13,%r14,52(%r15) # restore registers r13 & 14
584 40038e: 07 fe br %r14 # return
587 Comments on the function main
588 -----------------------------
589 1) The compiler did this function optimally ( 8-) )
591 Literal pool for main.
592 400390: ff ff ff ec .long 0xffffffec
593 main(int argc,char *argv[])
594 { # Function prologue below
595 400394: 90 bf f0 2c stm %r11,%r15,44(%r15) # Save necessary registers
596 400398: 18 0f lr %r0,%r15 # copy stack pointer to r0
597 40039a: a7 fa ff a0 ahi %r15,-96 # Make area for callee saving
598 40039e: 0d d0 basr %r13,%r0 # Set up r13 to point to
599 4003a0: a7 da ff f0 ahi %r13,-16 # literal pool
600 4003a4: 50 00 f0 00 st %r0,0(%r15) # Save backchain
602 return(test(5)); # Main Program Below
603 4003a8: 58 e0 d0 00 l %r14,0(%r13) # load relative address of test from
605 4003ac: a7 28 00 05 lhi %r2,5 # Set first parameter to 5
606 4003b0: 4d ee d0 00 bas %r14,0(%r14,%r13) # jump to test setting r14 as return
607 # address using branch & save instruction.
609 # Function Epilogue below
610 4003b4: 98 bf f0 8c lm %r11,%r15,140(%r15)# Restore necessary registers.
611 4003b8: 07 fe br %r14 # return to do program exit
618 main(int argc,char *argv[])
620 4004fc: 90 7f f0 1c stm %r7,%r15,28(%r15)
621 400500: a7 d5 00 04 bras %r13,400508 <main+0xc>
622 400504: 00 40 04 f4 .long 0x004004f4
623 # compiler now puts constant pool in code to so it saves an instruction
624 400508: 18 0f lr %r0,%r15
625 40050a: a7 fa ff a0 ahi %r15,-96
626 40050e: 50 00 f0 00 st %r0,0(%r15)
628 400512: 58 10 d0 00 l %r1,0(%r13)
629 400516: a7 28 00 05 lhi %r2,5
630 40051a: 0d e1 basr %r14,%r1
631 # compiler adds 1 extra instruction to epilogue this is done to
632 # avoid processor pipeline stalls owing to data dependencies on g5 &
633 # above as register 14 in the old code was needed directly after being loaded
634 # by the lm %r11,%r15,140(%r15) for the br %14.
635 40051c: 58 40 f0 98 l %r4,152(%r15)
636 400520: 98 7f f0 7c lm %r7,%r15,124(%r15)
641 Hartmut ( our compiler developer ) also has been threatening to take out the
642 stack backchain in optimised code as this also causes pipeline stalls, you
645 64 bit z/Architecture code disassembly
646 --------------------------------------
648 If you understand the stuff above you'll understand the stuff
649 below too so I'll avoid repeating myself & just say that
650 some of the instructions have g's on the end of them to indicate
651 they are 64 bit & the stack offsets are a bigger,
652 the only other difference you'll find between 32 & 64 bit is that
653 we now use f4 & f6 for floating point arguments on 64 bit.
654 00000000800005b0 <test>:
658 800005b0: a7 2a 00 05 ahi %r2,5
659 800005b4: b9 14 00 22 lgfr %r2,%r2 # downcast to integer
660 800005b8: 07 fe br %r14
661 800005ba: 07 07 bcr 0,%r7
666 00000000800005bc <main>:
667 main(int argc,char *argv[])
669 800005bc: eb bf f0 58 00 24 stmg %r11,%r15,88(%r15)
670 800005c2: b9 04 00 1f lgr %r1,%r15
671 800005c6: a7 fb ff 60 aghi %r15,-160
672 800005ca: e3 10 f0 00 00 24 stg %r1,0(%r15)
674 800005d0: a7 29 00 05 lghi %r2,5
675 # brasl allows jumps > 64k & is overkill here bras would do fune
676 800005d4: c0 e5 ff ff ff ee brasl %r14,800005b0 <test>
677 800005da: e3 40 f1 10 00 04 lg %r4,272(%r15)
678 800005e0: eb bf f0 f8 00 04 lmg %r11,%r15,248(%r15)
679 800005e6: 07 f4 br %r4
684 Compiling programs for debugging on Linux for s/390 & z/Architecture
685 ====================================================================
686 -gdwarf-2 now works it should be considered the default debugging
687 format for s/390 & z/Architecture as it is more reliable for debugging
688 shared libraries, normal -g debugging works much better now
689 Thanks to the IBM java compiler developers bug reports.
691 This is typically done adding/appending the flags -g or -gdwarf-2 to the
692 CFLAGS & LDFLAGS variables Makefile of the program concerned.
694 If using gdb & you would like accurate displays of registers &
695 stack traces compile without optimisation i.e make sure
696 that there is no -O2 or similar on the CFLAGS line of the Makefile &
697 the emitted gcc commands, obviously this will produce worse code
698 ( not advisable for shipment ) but it is an aid to the debugging process.
700 This aids debugging because the compiler will copy parameters passed in
701 in registers onto the stack so backtracing & looking at passed in
702 parameters will work, however some larger programs which use inline functions
703 will not compile without optimisation.
705 Debugging with optimisation has since much improved after fixing
706 some bugs, please make sure you are using gdb-5.0 or later developed
709 Figuring out gcc compile errors
710 ===============================
711 If you are getting a lot of syntax errors compiling a program & the problem
712 isn't blatantly obvious from the source.
713 It often helps to just preprocess the file, this is done with the -E
715 What this does is that it runs through the very first phase of compilation
716 ( compilation in gcc is done in several stages & gcc calls many programs to
717 achieve its end result ) with the -E option gcc just calls the gcc preprocessor (cpp).
718 The c preprocessor does the following, it joins all the files #included together
719 recursively ( #include files can #include other files ) & also the c file you wish to compile.
720 It puts a fully qualified path of the #included files in a comment & it
721 does macro expansion.
722 This is useful for debugging because
723 1) You can double check whether the files you expect to be included are the ones
724 that are being included ( e.g. double check that you aren't going to the i386 asm directory ).
725 2) Check that macro definitions aren't clashing with typedefs,
726 3) Check that definitions aren't being used before they are being included.
727 4) Helps put the line emitting the error under the microscope if it contains macros.
729 For convenience the Linux kernel's makefile will do preprocessing automatically for you
730 by suffixing the file you want built with .i ( instead of .o )
733 from the linux directory type
734 make arch/s390/kernel/signal.i
737 s390-gcc -D__KERNEL__ -I/home1/barrow/linux/include -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer
738 -fno-strict-aliasing -D__SMP__ -pipe -fno-strength-reduce -E arch/s390/kernel/signal.c
739 > arch/s390/kernel/signal.i
741 Now look at signal.i you should see something like.
744 # 1 "/home1/barrow/linux/include/asm/types.h" 1
745 typedef unsigned short umode_t;
746 typedef __signed__ char __s8;
747 typedef unsigned char __u8;
748 typedef __signed__ short __s16;
749 typedef unsigned short __u16;
751 If instead you are getting errors further down e.g.
752 unknown instruction:2515 "move.l" or better still unknown instruction:2515
753 "Fixme not implemented yet, call Martin" you are probably are attempting to compile some code
754 meant for another architecture or code that is simply not implemented, with a fixme statement
755 stuck into the inline assembly code so that the author of the file now knows he has work to do.
756 To look at the assembly emitted by gcc just before it is about to call gas ( the gnu assembler )
758 Again for your convenience the Linux kernel's Makefile will hold your hand &
759 do all this donkey work for you also by building the file with the .s suffix.
761 from the Linux directory type
762 make arch/s390/kernel/signal.s
764 s390-gcc -D__KERNEL__ -I/home1/barrow/linux/include -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer
765 -fno-strict-aliasing -D__SMP__ -pipe -fno-strength-reduce -S arch/s390/kernel/signal.c
766 -o arch/s390/kernel/signal.s
769 This will output something like, ( please note the constant pool & the useful comments
770 in the prologue to give you a hand at interpreting it ).
773 .string "misaligned (__u16 *) in __xchg\n"
775 .string "misaligned (__u32 *) in __xchg\n"
776 .L$PG1: # Pool sys_sigsuspend
782 .long schedule-.L$PG1
784 .long do_signal-.L$PG1
786 .globl sys_sigsuspend
787 .type sys_sigsuspend,@function
792 # need frame pointer 0
795 # incoming args (stack) 0
796 # function length 168
801 .L$CO1: AHI 13,.L$PG1-.L$CO1
804 N 5,.LC192-.L$PG1(13)
806 Adding -g to the above output makes the output even more useful
808 make CC:="s390-gcc -g" kernel/sched.s
811 s390-gcc -g -D__KERNEL__ -I/home/barrow/linux-2.3/include -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer -fno-strict-aliasing -pipe -fno-strength-reduce -S kernel/sched.c -o kernel/sched.s
813 also outputs stabs ( debugger ) info, from this info you can find out the
814 offsets & sizes of various elements in structures.
815 e.g. the stab for the structure
817 unsigned long rlim_cur;
818 unsigned long rlim_max;
821 .stabs "rlimit:T(151,2)=s8rlim_cur:(0,5),0,32;rlim_max:(0,5),32,32;;",128,0,0,0
822 from this stab you can see that
823 rlimit_cur starts at bit offset 0 & is 32 bits in size
824 rlimit_max starts at bit offset 32 & is 32 bits in size.
832 This is a tool with many options the most useful being ( if compiled with -g).
833 objdump --source <victim program or object file> > <victims debug listing >
836 The whole kernel can be compiled like this ( Doing this will make a 17MB kernel
837 & a 200 MB listing ) however you have to strip it before building the image
838 using the strip command to make it a more reasonable size to boot it.
840 A source/assembly mixed dump of the kernel can be done with the line
841 objdump --source vmlinux > vmlinux.lst
842 Also, if the file isn't compiled -g, this will output as much debugging information
843 as it can (e.g. function names). This is very slow as it spends lots
844 of time searching for debugging info. The following self explanatory line should be used
845 instead if the code isn't compiled -g, as it is much faster:
846 objdump --disassemble-all --syms vmlinux > vmlinux.lst
848 As hard drive space is valuable most of us use the following approach.
849 1) Look at the emitted psw on the console to find the crash address in the kernel.
850 2) Look at the file System.map ( in the linux directory ) produced when building
851 the kernel to find the closest address less than the current PSW to find the
853 3) use grep or similar to search the source tree looking for the source file
854 with this function if you don't know where it is.
855 4) rebuild this object file with -g on, as an example suppose the file was
856 ( /arch/s390/kernel/signal.o )
857 5) Assuming the file with the erroneous function is signal.c Move to the base of the
859 6) rm /arch/s390/kernel/signal.o
860 7) make /arch/s390/kernel/signal.o
861 8) watch the gcc command line emitted
862 9) type it in again or alternatively cut & paste it on the console adding the -g option.
863 10) objdump --source arch/s390/kernel/signal.o > signal.lst
864 This will output the source & the assembly intermixed, as the snippet below shows
865 This will unfortunately output addresses which aren't the same
866 as the kernel ones you should be able to get around the mental arithmetic
867 by playing with the --adjust-vma parameter to objdump.
872 static inline void spin_lock(spinlock_t *lp)
875 a2: a7 3a 03 bc ahi %r3,956
876 __asm__ __volatile(" lhi 1,-1\n"
877 a6: a7 18 ff ff lhi %r1,-1
878 aa: 1f 00 slr %r0,%r0
879 ac: ba 01 30 00 cs %r0,%r1,0(%r3)
880 b0: a7 44 ff fd jm aa <sys_sigsuspend+0x2e>
881 saveset = current->blocked;
882 b4: d2 07 f0 68 mvc 104(8,%r15),972(%r4)
884 return (set->sig[0] & mask) != 0;
887 6) If debugging under VM go down to that section in the document for more info.
890 I now have a tool which takes the pain out of --adjust-vma
891 & you are able to do something like
892 make /arch/s390/kernel/traps.lst
893 & it automatically generates the correctly relocated entries for
894 the text segment in traps.lst.
895 This tool is now standard in linux distro's in scripts/makelst
900 A. It is a tool for intercepting calls to the kernel & logging them
901 to a file & on the screen.
904 A. You can use it to find out what files a particular program opens.
910 If you wanted to know does ping work but didn't have the source
911 strace ping -c 1 127.0.0.1
912 & then look at the man pages for each of the syscalls below,
913 ( In fact this is sometimes easier than looking at some spaghetti
914 source which conditionally compiles for several architectures ).
915 Not everything that it throws out needs to make sense immediately.
917 Just looking quickly you can see that it is making up a RAW socket
918 for the ICMP protocol.
919 Doing an alarm(10) for a 10 second timeout
920 & doing a gettimeofday call before & after each read to see
921 how long the replies took, & writing some text to stdout so the user
922 has an idea what is going on.
924 socket(PF_INET, SOCK_RAW, IPPROTO_ICMP) = 3
927 stat("/usr/share/locale/C/libc.cat", 0xbffff134) = -1 ENOENT (No such file or directory)
928 stat("/usr/share/locale/libc/C", 0xbffff134) = -1 ENOENT (No such file or directory)
929 stat("/usr/local/share/locale/C/libc.cat", 0xbffff134) = -1 ENOENT (No such file or directory)
931 setsockopt(3, SOL_SOCKET, SO_BROADCAST, [1], 4) = 0
932 setsockopt(3, SOL_SOCKET, SO_RCVBUF, [49152], 4) = 0
933 fstat(1, {st_mode=S_IFCHR|0620, st_rdev=makedev(3, 1), ...}) = 0
934 mmap(0, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x40008000
935 ioctl(1, TCGETS, {B9600 opost isig icanon echo ...}) = 0
936 write(1, "PING 127.0.0.1 (127.0.0.1): 56 d"..., 42PING 127.0.0.1 (127.0.0.1): 56 data bytes
938 sigaction(SIGINT, {0x8049ba0, [], SA_RESTART}, {SIG_DFL}) = 0
939 sigaction(SIGALRM, {0x8049600, [], SA_RESTART}, {SIG_DFL}) = 0
940 gettimeofday({948904719, 138951}, NULL) = 0
941 sendto(3, "\10\0D\201a\1\0\0\17#\2178\307\36"..., 64, 0, {sin_family=AF_INET,
942 sin_port=htons(0), sin_addr=inet_addr("127.0.0.1")}, 16) = 64
943 sigaction(SIGALRM, {0x8049600, [], SA_RESTART}, {0x8049600, [], SA_RESTART}) = 0
944 sigaction(SIGALRM, {0x8049ba0, [], SA_RESTART}, {0x8049600, [], SA_RESTART}) = 0
946 recvfrom(3, "E\0\0T\0005\0\0@\1|r\177\0\0\1\177"..., 192, 0,
947 {sin_family=AF_INET, sin_port=htons(50882), sin_addr=inet_addr("127.0.0.1")}, [16]) = 84
948 gettimeofday({948904719, 160224}, NULL) = 0
949 recvfrom(3, "E\0\0T\0006\0\0\377\1\275p\177\0"..., 192, 0,
950 {sin_family=AF_INET, sin_port=htons(50882), sin_addr=inet_addr("127.0.0.1")}, [16]) = 84
951 gettimeofday({948904719, 166952}, NULL) = 0
952 write(1, "64 bytes from 127.0.0.1: icmp_se"...,
953 5764 bytes from 127.0.0.1: icmp_seq=0 ttl=255 time=28.0 ms
957 strace passwd 2>&1 | grep open
958 produces the following output
959 open("/etc/ld.so.cache", O_RDONLY) = 3
960 open("/opt/kde/lib/libc.so.5", O_RDONLY) = -1 ENOENT (No such file or directory)
961 open("/lib/libc.so.5", O_RDONLY) = 3
962 open("/dev", O_RDONLY) = 3
963 open("/var/run/utmp", O_RDONLY) = 3
964 open("/etc/passwd", O_RDONLY) = 3
965 open("/etc/shadow", O_RDONLY) = 3
966 open("/etc/login.defs", O_RDONLY) = 4
967 open("/dev/tty", O_RDONLY) = 4
969 The 2>&1 is done to redirect stderr to stdout & grep is then filtering this input
970 through the pipe for each line containing the string open.
975 Getting sophisticated
976 telnetd crashes & I don't know why
980 1) Replace the following line in /etc/inetd.conf
981 telnet stream tcp nowait root /usr/sbin/in.telnetd -h
983 telnet stream tcp nowait root /blah
985 2) Create the file /blah with the following contents to start tracing telnetd
987 /usr/bin/strace -o/t1 -f /usr/sbin/in.telnetd -h
988 3) chmod 700 /blah to make it executable only to root
991 or ps aux | grep inetd
992 get inetd's process id
993 & kill -HUP inetd to restart it.
997 -o is used to tell strace to output to a file in our case t1 in the root directory
998 -f is to follow children i.e.
999 e.g in our case above telnetd will start the login process & subsequently a shell like bash.
1000 You will be able to tell which is which from the process ID's listed on the left hand side
1001 of the strace output.
1002 -p<pid> will tell strace to attach to a running process, yup this can be done provided
1003 it isn't being traced or debugged already & you have enough privileges,
1004 the reason 2 processes cannot trace or debug the same program is that strace
1005 becomes the parent process of the one being debugged & processes ( unlike people )
1006 can have only one parent.
1009 However the file /t1 will get big quite quickly
1010 to test it telnet 127.0.0.1
1012 now look at what files in.telnetd execve'd
1013 413 execve("/usr/sbin/in.telnetd", ["/usr/sbin/in.telnetd", "-h"], [/* 17 vars */]) = 0
1014 414 execve("/bin/login", ["/bin/login", "-h", "localhost", "-p"], [/* 2 vars */]) = 0
1021 If the program is not very interactive ( i.e. not much keyboard input )
1022 & is crashing in one architecture but not in another you can do
1023 an strace of both programs under as identical a scenario as you can
1024 on both architectures outputting to a file then.
1025 do a diff of the two traces using the diff program
1027 diff output1 output2
1028 & maybe you'll be able to see where the call paths differed, this
1029 is possibly near the cause of the crash.
1033 Look at man pages for strace & the various syscalls
1034 e.g. man strace, man alarm, man socket.
1037 Performance Debugging
1038 =====================
1039 gcc is capable of compiling in profiling code just add the -p option
1040 to the CFLAGS, this obviously affects program size & performance.
1041 This can be used by the gprof gnu profiling tool or the
1042 gcov the gnu code coverage tool ( code coverage is a means of testing
1043 code quality by checking if all the code in an executable in exercised by
1047 Using top to find out where processes are sleeping in the kernel
1048 ----------------------------------------------------------------
1049 To do this copy the System.map from the root directory where
1050 the linux kernel was built to the /boot directory on your
1054 You should see a new field called WCHAN which
1055 tells you where each process is sleeping here is a typical output.
1057 6:59pm up 41 min, 1 user, load average: 0.00, 0.00, 0.00
1058 28 processes: 27 sleeping, 1 running, 0 zombie, 0 stopped
1059 CPU states: 0.0% user, 0.1% system, 0.0% nice, 99.8% idle
1060 Mem: 254900K av, 45976K used, 208924K free, 0K shrd, 28636K buff
1061 Swap: 0K av, 0K used, 0K free 8620K cached
1063 PID USER PRI NI SIZE RSS SHARE WCHAN STAT LIB %CPU %MEM TIME COMMAND
1064 750 root 12 0 848 848 700 do_select S 0 0.1 0.3 0:00 in.telnetd
1065 767 root 16 0 1140 1140 964 R 0 0.1 0.4 0:00 top
1066 1 root 8 0 212 212 180 do_select S 0 0.0 0.0 0:00 init
1067 2 root 9 0 0 0 0 down_inte SW 0 0.0 0.0 0:00 kmcheck
1071 Another related command is the time command which gives you an indication
1072 of where a process is spending the majority of its time.
1085 Addresses & values in the VM debugger are always hex never decimal
1086 Address ranges are of the format <HexValue1>-<HexValue2> or <HexValue1>.<HexValue2>
1087 e.g. The address range 0x2000 to 0x3000 can be described as 2000-3000 or 2000.1000
1089 The VM Debugger is case insensitive.
1091 VM's strengths are usually other debuggers weaknesses you can get at any resource
1092 no matter how sensitive e.g. memory management resources,change address translation
1093 in the PSW. For kernel hacking you will reap dividends if you get good at it.
1095 The VM Debugger displays operators but not operands, probably because some
1096 of it was written when memory was expensive & the programmer was probably proud that
1097 it fitted into 2k of memory & the programmers & didn't want to shock hardcore VM'ers by
1098 changing the interface :-), also the debugger displays useful information on the same line &
1099 the author of the code probably felt that it was a good idea not to go over
1100 the 80 columns on the screen.
1102 As some of you are probably in a panic now this isn't as unintuitive as it may seem
1103 as the 390 instructions are easy to decode mentally & you can make a good guess at a lot
1104 of them as all the operands are nibble ( half byte aligned ) & if you have an objdump listing
1105 also it is quite easy to follow, if you don't have an objdump listing keep a copy of
1106 the s/390 Reference Summary & look at between pages 2 & 7 or alternatively the
1107 s/390 principles of operation.
1108 e.g. even I can guess that
1109 0001AFF8' LR 180F CC 0
1110 is a ( load register ) lr r0,r15
1112 Also it is very easy to tell the length of a 390 instruction from the 2 most significant
1113 bits in the instruction ( not that this info is really useful except if you are trying to
1114 make sense of a hexdump of code ).
1116 Bits Instruction Length
1117 ------------------------------------------
1126 The debugger also displays other useful info on the same line such as the
1127 addresses being operated on destination addresses of branches & condition codes.
1129 00019736' AHI A7DAFF0E CC 1
1130 000198BA' BRC A7840004 -> 000198C2' CC 0
1131 000198CE' STM 900EF068 >> 0FA95E78 CC 2
1135 Useful VM debugger commands
1136 ---------------------------
1138 I suppose I'd better mention this before I start
1139 to list the current active traces do
1141 there can be a maximum of 255 of these per set
1142 ( more about trace sets later ).
1143 To stop traces issue a
1145 To delete a particular breakpoint issue
1146 TR DEL <breakpoint number>
1148 The PA1 key drops to CP mode so you can issue debugger commands,
1149 Doing alt c (on my 3270 console at least ) clears the screen.
1150 hitting b <enter> comes back to the running operating system
1151 from cp mode ( in our case linux ).
1152 It is typically useful to add shortcuts to your profile.exec file
1153 if you have one ( this is roughly equivalent to autoexec.bat in DOS ).
1154 file here are a few from mine.
1155 /* this gives me command history on issuing f12 */
1157 /* this continues */
1159 /* goes to trace set a */
1160 set pf1 imm tr goto a
1161 /* goes to trace set b */
1162 set pf2 imm tr goto b
1163 /* goes to trace set c */
1164 set pf3 imm tr goto c
1170 Setting a simple breakpoint
1172 To debug a particular function try
1173 TR I R <function address range>
1174 TR I on its own will single step.
1175 TR I DATA <MNEMONIC> <OPTIONAL RANGE> will trace for particular mnemonics
1177 TR I DATA 4D R 0197BC.4000
1178 will trace for BAS'es ( opcode 4D ) in the range 0197BC.4000
1179 if you were inclined you could add traces for all branch instructions &
1180 suffix them with the run prefix so you would have a backtrace on screen
1181 when a program crashes.
1182 TR BR <INTO OR FROM> will trace branches into or out of an address.
1184 TR BR INTO 0 is often quite useful if a program is getting awkward & deciding
1185 to branch to 0 & crashing as this will stop at the address before in jumps to 0.
1186 TR I R <address range> RUN cmd d g
1187 single steps a range of addresses but stays running &
1188 displays the gprs on each step.
1192 Displaying & modifying Registers
1193 --------------------------------
1194 D G will display all the gprs
1195 Adding a extra G to all the commands is necessary to access the full 64 bit
1196 content in VM on z/Architecture obviously this isn't required for access registers
1197 as these are still 32 bit.
1198 e.g. DGG instead of DG
1199 D X will display all the control registers
1200 D AR will display all the access registers
1201 D AR4-7 will display access registers 4 to 7
1202 CPU ALL D G will display the GRPS of all CPUS in the configuration
1203 D PSW will display the current PSW
1204 st PSW 2000 will put the value 2000 into the PSW &
1205 cause crash your machine.
1206 D PREFIX displays the prefix offset
1211 To display memory mapped using the current PSW's mapping try
1213 To make VM display a message each time it hits a particular address & continue try
1214 D I<range> will disassemble/display a range of instructions.
1215 ST addr 32 bit word will store a 32 bit aligned address
1216 D T<range> will display the EBCDIC in an address ( if you are that way inclined )
1217 D R<range> will display real addresses ( without DAT ) but with prefixing.
1218 There are other complex options to display if you need to get at say home space
1219 but are in primary space the easiest thing to do is to temporarily
1220 modify the PSW to the other addressing mode, display the stuff & then
1227 If you want to issue a debugger command without halting your virtual machine with the
1228 PA1 key try prefixing the command with #CP e.g.
1230 also suffixing most debugger commands with RUN will cause them not
1231 to stop just display the mnemonic at the current instruction on the console.
1232 If you have several breakpoints you want to put into your program &
1233 you get fed up of cross referencing with System.map
1234 you can do the following trick for several symbols.
1235 grep do_signal System.map
1236 which emits the following among other things
1237 0001f4e0 T do_signal
1240 TR I PSWA 0001f4e0 cmd msg * do_signal
1241 This sends a message to your own console each time do_signal is entered.
1242 ( As an aside I wrote a perl script once which automatically generated a REXX
1243 script with breakpoints on every kernel procedure, this isn't a good idea
1244 because there are thousands of these routines & VM can only set 255 breakpoints
1245 at a time so you nearly had to spend as long pruning the file down as you would
1246 entering the msg's by hand ),however, the trick might be useful for a single object file.
1247 On linux'es 3270 emulator x3270 there is a very useful option under the file ment
1248 Save Screens In File this is very good of keeping a copy of traces.
1250 From CMS help <command name> will give you online help on a particular command.
1254 Also CP has a file called profile.exec which automatically gets called
1255 on startup of CMS ( like autoexec.bat ), keeping on a DOS analogy session
1256 CP has a feature similar to doskey, it may be useful for you to
1257 use profile.exec to define some keystrokes.
1260 This does a single step in VM on pressing F8.
1262 This sets up the ^ key.
1263 which can be used for ^c (ctrl-c),^z (ctrl-z) which can't be typed directly into some 3270 consoles.
1265 This types the starting keystrokes for a sysrq see SysRq below.
1267 This retrieves command history on pressing F12.
1270 Sometimes in VM the display is set up to scroll automatically this
1271 can be very annoying if there are messages you wish to look at
1274 This will nearly stop automatic screen updates, however it will
1275 cause a denial of service if lots of messages go to the 3270 console,
1276 so it would be foolish to use this as the default on a production machine.
1279 Tracing particular processes
1280 ----------------------------
1281 The kernel's text segment is intentionally at an address in memory that it will
1282 very seldom collide with text segments of user programs ( thanks Martin ),
1283 this simplifies debugging the kernel.
1284 However it is quite common for user processes to have addresses which collide
1285 this can make debugging a particular process under VM painful under normal
1286 circumstances as the process may change when doing a
1287 TR I R <address range>.
1288 Thankfully after reading VM's online help I figured out how to debug
1289 I particular process.
1291 Your first problem is to find the STD ( segment table designation )
1292 of the program you wish to debug.
1293 There are several ways you can do this here are a few
1294 1) objdump --syms <program to be debugged> | grep main
1295 To get the address of main in the program.
1296 tr i pswa <address of main>
1297 Start the program, if VM drops to CP on what looks like the entry
1298 point of the main function this is most likely the process you wish to debug.
1299 Now do a D X13 or D XG13 on z/Architecture.
1300 On 31 bit the STD is bits 1-19 ( the STO segment table origin )
1301 & 25-31 ( the STL segment table length ) of CR13.
1303 TR I R STD <CR13's value> 0.7fffffff
1305 TR I R STD 8F32E1FF 0.7fffffff
1306 Another very useful variation is
1307 TR STORE INTO STD <CR13's value> <address range>
1308 for finding out when a particular variable changes.
1310 An alternative way of finding the STD of a currently running process
1311 is to do the following, ( this method is more complex but
1312 could be quite convenient if you aren't updating the kernel much &
1313 so your kernel structures will stay constant for a reasonable period of
1316 grep task /proc/<pid>/status
1317 from this you should see something like
1318 task: 0f160000 ksp: 0f161de8 pt_regs: 0f161f68
1319 This now gives you a pointer to the task structure.
1320 Now make CC:="s390-gcc -g" kernel/sched.s
1321 To get the task_struct stabinfo.
1322 ( task_struct is defined in include/linux/sched.h ).
1323 Now we want to look at
1324 task->active_mm->pgd
1325 on my machine the active_mm in the task structure stab is
1326 active_mm:(4,12),672,32
1327 its offset is 672/8=84=0x54
1328 the pgd member in the mm_struct stab is
1329 pgd:(4,6)=*(29,5),96,32
1330 so its offset is 96/8=12=0xc
1333 hexdump -s 0xf160054 /dev/mem | more
1334 i.e. task_struct+active_mm offset
1335 to look at the active_mm member
1336 f160054 0fee cc60 0019 e334 0000 0000 0000 0011
1337 hexdump -s 0x0feecc6c /dev/mem | more
1338 i.e. active_mm+pgd offset
1339 feecc6c 0f2c 0000 0000 0001 0000 0001 0000 0010
1340 we get something like
1342 TR I R STD <pgd|0x7f> 0.7fffffff
1343 i.e. the 0x7f is added because the pgd only
1344 gives the page table origin & we need to set the low bits
1345 to the maximum possible segment table length.
1346 TR I R STD 0f2c007f 0.7fffffff
1347 on z/Architecture you'll probably need to do
1348 TR I R STD <pgd|0x7> 0.ffffffffffffffff
1349 to set the TableType to 0x1 & the Table length to 3.
1353 Tracing Program Exceptions
1354 --------------------------
1355 If you get a crash which says something like
1356 illegal operation or specification exception followed by a register dump
1357 You can restart linux & trace these using the tr prog <range or value> trace option.
1361 The most common ones you will normally be tracing for is
1362 1=operation exception
1363 2=privileged operation exception
1364 4=protection exception
1365 5=addressing exception
1366 6=specification exception
1367 10=segment translation exception
1368 11=page translation exception
1370 The full list of these is on page 22 of the current s/390 Reference Summary.
1372 tr prog 10 will trace segment translation exceptions.
1373 tr prog on its own will trace all program interruption codes.
1377 On starting VM you are initially in the INITIAL trace set.
1378 You can do a Q TR to verify this.
1379 If you have a complex tracing situation where you wish to wait for instance
1380 till a driver is open before you start tracing IO, but know in your
1381 heart that you are going to have to make several runs through the code till you
1382 have a clue whats going on.
1385 TR I PSWA <Driver open address>
1386 hit b to continue till breakpoint
1387 reach the breakpoint
1390 TR IO 7c08-7c09 inst int run
1391 or whatever the IO channels you wish to trace are & hit b
1393 To got back to the initial trace set do
1395 & the TR I PSWA <Driver open address> will be the only active breakpoint again.
1398 Tracing linux syscalls under VM
1399 -------------------------------
1400 Syscalls are implemented on Linux for S390 by the Supervisor call instruction (SVC) there 256
1401 possibilities of these as the instruction is made up of a 0xA opcode & the second byte being
1402 the syscall number. They are traced using the simple command.
1403 TR SVC <Optional value or range>
1404 the syscalls are defined in linux/arch/s390/include/asm/unistd.h
1405 e.g. to trace all file opens just do
1406 TR SVC 5 ( as this is the syscall number of open )
1409 SMP Specific commands
1410 ---------------------
1411 To find out how many cpus you have
1412 Q CPUS displays all the CPU's available to your virtual machine
1413 To find the cpu that the current cpu VM debugger commands are being directed at do
1414 Q CPU to change the current cpu VM debugger commands are being directed at do
1415 CPU <desired cpu no>
1417 On a SMP guest issue a command to all CPUs try prefixing the command with cpu all.
1418 To issue a command to a particular cpu try cpu <cpu number> e.g.
1419 CPU 01 TR I R 2000.3000
1420 If you are running on a guest with several cpus & you have a IO related problem
1421 & cannot follow the flow of code but you know it isn't smp related.
1422 from the bash prompt issue
1423 shutdown -h now or halt.
1424 do a Q CPUS to find out how many cpus you have
1425 detach each one of them from cp except cpu 0
1427 DETACH CPU 01-(number of cpus in configuration)
1429 TR SIGP will trace inter processor signal processor instructions.
1430 DEFINE CPU 01-(number in configuration)
1431 will get your guests cpus back.
1434 Help for displaying ascii textstrings
1435 -------------------------------------
1436 On the very latest VM Nucleus'es VM can now display ascii
1437 ( thanks Neale for the hint ) by doing
1444 Under older VM debuggers ( I love EBDIC too ) you can use this little program I wrote which
1445 will convert a command line of hex digits to ascii text which can be compiled under linux &
1446 you can copy the hex digits from your x3270 terminal to your xterm if you are debugging
1449 This is quite useful when looking at a parameter passed in as a text string
1450 under VM ( unless you are good at decoding ASCII in your head ).
1452 e.g. consider tracing an open syscall
1454 We have stopped at a breakpoint
1455 000151B0' SVC 0A05 -> 0001909A' CC 0
1457 D 20.8 to check the SVC old psw in the prefix area & see was it from userspace
1458 ( for the layout of the prefix area consult P18 of the s/390 390 Reference Summary
1459 if you have it available ).
1460 V00000020 070C2000 800151B2
1461 The problem state bit wasn't set & it's also too early in the boot sequence
1462 for it to be a userspace SVC if it was we would have to temporarily switch the
1463 psw to user space addressing so we could get at the first parameter of the open in
1468 Now display what gpr2 is pointing to
1470 V00014CB4 2F646576 2F636F6E 736F6C65 00001BF5
1471 V00014CC4 FC00014C B4001001 E0001000 B8070707
1472 Now copy the text till the first 00 hex ( which is the end of the string
1473 to an xterm & do hex2ascii on it.
1474 hex2ascii 2F646576 2F636F6E 736F6C65 00
1476 Decoded Hex:=/ d e v / c o n s o l e 0x00
1477 We were opening the console device,
1479 You can compile the code below yourself for practice :-),
1482 * a useful little tool for converting a hexadecimal command line to ascii
1484 * Author(s): Denis Joseph Barrow (djbarrow@de.ibm.com,barrow_dj@yahoo.com)
1485 * (C) 2000 IBM Deutschland Entwicklung GmbH, IBM Corporation.
1489 int main(int argc,char *argv[])
1491 int cnt1,cnt2,len,toggle=0;
1493 unsigned char c,hex;
1495 if(argc>1&&(strcmp(argv[1],"-a")==0))
1497 printf("Decoded Hex:=");
1498 for(cnt1=startcnt;cnt1<argc;cnt1++)
1500 len=strlen(argv[cnt1]);
1501 for(cnt2=0;cnt2<len;cnt2++)
1521 printf("0x%02X ",(int)hex);
1542 Stack tracing under VM
1543 ----------------------
1547 Here are the tricks I use 9 out of 10 times it works pretty well,
1549 When your backchain reaches a dead end
1550 --------------------------------------
1551 This can happen when an exception happens in the kernel & the kernel is entered twice
1552 if you reach the NULL pointer at the end of the back chain you should be
1553 able to sniff further back if you follow the following tricks.
1554 1) A kernel address should be easy to recognise since it is in
1555 primary space & the problem state bit isn't set & also
1556 The Hi bit of the address is set.
1557 2) Another backchain should also be easy to recognise since it is an
1558 address pointing to another address approximately 100 bytes or 0x70 hex
1559 behind the current stackpointer.
1562 Here is some practice.
1563 boot the kernel & hit PA1 at some random time
1564 d g to display the gprs, this should display something like
1565 GPR 0 = 00000001 00156018 0014359C 00000000
1566 GPR 4 = 00000001 001B8888 000003E0 00000000
1567 GPR 8 = 00100080 00100084 00000000 000FE000
1568 GPR 12 = 00010400 8001B2DC 8001B36A 000FFED8
1569 Note that GPR14 is a return address but as we are real men we are going to
1571 display 0x40 bytes after the stack pointer.
1573 V000FFED8 000FFF38 8001B838 80014C8E 000FFF38
1574 V000FFEE8 00000000 00000000 000003E0 00000000
1575 V000FFEF8 00100080 00100084 00000000 000FE000
1576 V000FFF08 00010400 8001B2DC 8001B36A 000FFED8
1579 Ah now look at whats in sp+56 (sp+0x38) this is 8001B36A our saved r14 if
1580 you look above at our stackframe & also agrees with GPR14.
1584 we now are taking the contents of SP to get our first backchain.
1586 V000FFF38 000FFFA0 00000000 00014995 00147094
1587 V000FFF48 00147090 001470A0 000003E0 00000000
1588 V000FFF58 00100080 00100084 00000000 001BF1D0
1589 V000FFF68 00010400 800149BA 80014CA6 000FFF38
1591 This displays a 2nd return address of 80014CA6
1593 now do d 000FFFA0.40 for our 3rd backchain
1595 V000FFFA0 04B52002 0001107F 00000000 00000000
1596 V000FFFB0 00000000 00000000 FF000000 0001107F
1597 V000FFFC0 00000000 00000000 00000000 00000000
1598 V000FFFD0 00010400 80010802 8001085A 000FFFA0
1601 our 3rd return address is 8001085A
1603 as the 04B52002 looks suspiciously like rubbish it is fair to assume that the kernel entry routines
1604 for the sake of optimisation don't set up a backchain.
1606 now look at System.map to see if the addresses make any sense.
1608 grep -i 0001b3 System.map
1609 outputs among other things
1612 is cpu_idle+0x66 ( quiet the cpu is asleep, don't wake it )
1615 grep -i 00014 System.map
1616 produces among other things
1617 00014a78 T start_kernel
1618 so 0014CA6 is start_kernel+some hex number I can't add in my head.
1620 grep -i 00108 System.map
1623 so 8001085A is _stext+0x5a
1625 Congrats you've done your first backchain.
1629 s/390 & z/Architecture IO Overview
1630 ==================================
1632 I am not going to give a course in 390 IO architecture as this would take me quite a
1633 while & I'm no expert. Instead I'll give a 390 IO architecture summary for Dummies if you have
1634 the s/390 principles of operation available read this instead. If nothing else you may find a few
1635 useful keywords in here & be able to use them on a web search engine like altavista to find
1636 more useful information.
1638 Unlike other bus architectures modern 390 systems do their IO using mostly
1639 fibre optics & devices such as tapes & disks can be shared between several mainframes,
1640 also S390 can support up to 65536 devices while a high end PC based system might be choking
1641 with around 64. Here is some of the common IO terminology
1644 This is the logical number most IO commands use to talk to an IO device there can be up to
1645 0x10000 (65536) of these in a configuration typically there is a few hundred. Under VM
1646 for simplicity they are allocated contiguously, however on the native hardware they are not
1647 they typically stay consistent between boots provided no new hardware is inserted or removed.
1648 Under Linux for 390 we use these as IRQ's & also when issuing an IO command (CLEAR SUBCHANNEL,
1649 HALT SUBCHANNEL,MODIFY SUBCHANNEL,RESUME SUBCHANNEL,START SUBCHANNEL,STORE SUBCHANNEL &
1650 TEST SUBCHANNEL ) we use this as the ID of the device we wish to talk to, the most
1651 important of these instructions are START SUBCHANNEL ( to start IO ), TEST SUBCHANNEL ( to check
1652 whether the IO completed successfully ), & HALT SUBCHANNEL ( to kill IO ), a subchannel
1653 can have up to 8 channel paths to a device this offers redundancy if one is not available.
1657 This number remains static & Is closely tied to the hardware, there are 65536 of these
1658 also they are made up of a CHPID ( Channel Path ID, the most significant 8 bits )
1659 & another lsb 8 bits. These remain static even if more devices are inserted or removed
1660 from the hardware, there is a 1 to 1 mapping between Subchannels & Device Numbers provided
1661 devices aren't inserted or removed.
1663 Channel Control Words:
1664 CCWS are linked lists of instructions initially pointed to by an operation request block (ORB),
1665 which is initially given to Start Subchannel (SSCH) command along with the subchannel number
1666 for the IO subsystem to process while the CPU continues executing normal code.
1667 These come in two flavours, Format 0 ( 24 bit for backward )
1668 compatibility & Format 1 ( 31 bit ). These are typically used to issue read & write
1669 ( & many other instructions ) they consist of a length field & an absolute address field.
1670 For each IO typically get 1 or 2 interrupts one for channel end ( primary status ) when the
1671 channel is idle & the second for device end ( secondary status ) sometimes you get both
1672 concurrently, you check how the IO went on by issuing a TEST SUBCHANNEL at each interrupt,
1673 from which you receive an Interruption response block (IRB). If you get channel & device end
1674 status in the IRB without channel checks etc. your IO probably went okay. If you didn't you
1675 probably need a doctor to examine the IRB & extended status word etc.
1676 If an error occurs, more sophisticated control units have a facility known as
1677 concurrent sense this means that if an error occurs Extended sense information will
1678 be presented in the Extended status word in the IRB if not you have to issue a
1679 subsequent SENSE CCW command after the test subchannel.
1682 TPI( Test pending interrupt) can also be used for polled IO but in multitasking multiprocessor
1683 systems it isn't recommended except for checking special cases ( i.e. non looping checks for
1686 Store Subchannel & Modify Subchannel can be used to examine & modify operating characteristics
1687 of a subchannel ( e.g. channel paths ).
1689 Other IO related Terms:
1690 Sysplex: S390's Clustering Technology
1691 QDIO: S390's new high speed IO architecture to support devices such as gigabit ethernet,
1692 this architecture is also designed to be forward compatible with up & coming 64 bit machines.
1697 Input Output Processors (IOP's) are responsible for communicating between
1698 the mainframe CPU's & the channel & relieve the mainframe CPU's from the
1699 burden of communicating with IO devices directly, this allows the CPU's to
1700 concentrate on data processing.
1702 IOP's can use one or more links ( known as channel paths ) to talk to each
1703 IO device. It first checks for path availability & chooses an available one,
1704 then starts ( & sometimes terminates IO ).
1705 There are two types of channel path: ESCON & the Parallel IO interface.
1707 IO devices are attached to control units, control units provide the
1708 logic to interface the channel paths & channel path IO protocols to
1709 the IO devices, they can be integrated with the devices or housed separately
1710 & often talk to several similar devices ( typical examples would be raid
1711 controllers or a control unit which connects to 1000 3270 terminals ).
1714 +---------------------------------------------------------------+
1715 | +-----+ +-----+ +-----+ +-----+ +----------+ +----------+ |
1716 | | CPU | | CPU | | CPU | | CPU | | Main | | Expanded | |
1717 | | | | | | | | | | Memory | | Storage | |
1718 | +-----+ +-----+ +-----+ +-----+ +----------+ +----------+ |
1719 |---------------------------------------------------------------+
1721 |---------------------------------------------------------------
1722 | C | C | C | C | C | C | C | C | C | C | C | C | C | C | C | C |
1723 ----------------------------------------------------------------
1725 || Bus & Tag Channel Path || ESCON
1726 || ====================== || Channel
1728 +----------+ +----------+ +----------+
1730 | CU | | CU | | CU |
1732 +----------+ +----------+ +----------+
1734 +----------+ +----------+ +----------+ +----------+ +----------+
1735 |I/O Device| |I/O Device| |I/O Device| |I/O Device| |I/O Device|
1736 +----------+ +----------+ +----------+ +----------+ +----------+
1737 CPU = Central Processing Unit
1742 The 390 IO systems come in 2 flavours the current 390 machines support both
1744 The Older 360 & 370 Interface,sometimes called the Parallel I/O interface,
1745 sometimes called Bus-and Tag & sometimes Original Equipment Manufacturers
1748 This byte wide Parallel channel path/bus has parity & data on the "Bus" cable
1749 & control lines on the "Tag" cable. These can operate in byte multiplex mode for
1750 sharing between several slow devices or burst mode & monopolize the channel for the
1751 whole burst. Up to 256 devices can be addressed on one of these cables. These cables are
1752 about one inch in diameter. The maximum unextended length supported by these cables is
1753 125 Meters but this can be extended up to 2km with a fibre optic channel extended
1754 such as a 3044. The maximum burst speed supported is 4.5 megabytes per second however
1755 some really old processors support only transfer rates of 3.0, 2.0 & 1.0 MB/sec.
1756 One of these paths can be daisy chained to up to 8 control units.
1759 ESCON if fibre optic it is also called FICON
1760 Was introduced by IBM in 1990. Has 2 fibre optic cables & uses either leds or lasers
1761 for communication at a signaling rate of up to 200 megabits/sec. As 10bits are transferred
1762 for every 8 bits info this drops to 160 megabits/sec & to 18.6 Megabytes/sec once
1763 control info & CRC are added. ESCON only operates in burst mode.
1765 ESCONs typical max cable length is 3km for the led version & 20km for the laser version
1766 known as XDF ( extended distance facility ). This can be further extended by using an
1767 ESCON director which triples the above mentioned ranges. Unlike Bus & Tag as ESCON is
1768 serial it uses a packet switching architecture the standard Bus & Tag control protocol
1769 is however present within the packets. Up to 256 devices can be attached to each control
1770 unit that uses one of these interfaces.
1772 Common 390 Devices include:
1773 Network adapters typically OSA2,3172's,2116's & OSA-E gigabit ethernet adapters,
1774 Consoles 3270 & 3215 ( a teletype emulated under linux for a line mode console ).
1775 DASD's direct access storage devices ( otherwise known as hard disks ).
1777 CTC ( Channel to Channel Adapters ),
1778 ESCON or Parallel Cables used as a very high speed serial link
1779 between 2 machines. We use 2 cables under linux to do a bi-directional serial link.
1782 Debugging IO on s/390 & z/Architecture under VM
1783 ===============================================
1785 Now we are ready to go on with IO tracing commands under VM
1787 A few self explanatory queries:
1790 Q DISK ( This command is CMS specific )
1798 Q OSA on my machine returns
1799 OSA 7C08 ON OSA 7C08 SUBCHANNEL = 0000
1800 OSA 7C09 ON OSA 7C09 SUBCHANNEL = 0001
1801 OSA 7C14 ON OSA 7C14 SUBCHANNEL = 0002
1802 OSA 7C15 ON OSA 7C15 SUBCHANNEL = 0003
1804 If you have a guest with certain privileges you may be able to see devices
1805 which don't belong to you. To avoid this, add the option V.
1809 Now using the device numbers returned by this command we will
1810 Trace the io starting up on the first device 7c08 & 7c09
1811 In our simplest case we can trace the
1813 like TR SSCH 7C08-7C09
1814 or the halt subchannels
1815 or TR HSCH 7C08-7C09
1816 MSCH's ,STSCH's I think you can guess the rest
1818 Ingo's favourite trick is tracing all the IO's & CCWS & spooling them into the reader of another
1819 VM guest so he can ftp the logfile back to his own machine.I'll do a small bit of this & give you
1820 a look at the output.
1822 1) Spool stdout to VM reader
1823 SP PRT TO (another vm guest ) or * for the local vm guest
1824 2) Fill the reader with the trace
1825 TR IO 7c08-7c09 INST INT CCW PRT RUN
1832 6) list reader contents
1834 7) copy it to linux4's minidisk
1835 RECEIVE / LOG TXT A1 ( replace
1837 filel & press F11 to look at it
1838 You should see something like:
1840 00020942' SSCH B2334000 0048813C CC 0 SCH 0000 DEV 7C08
1841 CPA 000FFDF0 PARM 00E2C9C4 KEY 0 FPI C0 LPM 80
1842 CCW 000FFDF0 E4200100 00487FE8 0000 E4240100 ........
1845 00020B0A' I/O DEV 7C08 -> 000197BC' SCH 0000 PARM 00E2C9C4
1846 00021628' TSCH B2354000 >> 00488164 CC 0 SCH 0000 DEV 7C08
1847 CCWA 000FFDF8 DEV STS 0C SCH STS 00 CNT 00EC
1848 KEY 0 FPI C0 CC 0 CTLS 4007
1849 00022238' STSCH B2344000 >> 00488108 CC 0 SCH 0000 DEV 7C08
1851 If you don't like messing up your readed ( because you possibly booted from it )
1852 you can alternatively spool it to another readers guest.
1855 Other common VM device related commands
1856 ---------------------------------------------
1857 These commands are listed only because they have
1858 been of use to me in the past & may be of use to
1859 you too. For more complete info on each of the commands
1860 use type HELP <command> from CMS.
1863 ATT <devno range> <guest>
1864 attach a device to guest * for your own guest
1865 READY <devno> cause VM to issue a fake interrupt.
1867 The VARY command is normally only available to VM administrators.
1868 VARY ON PATH <path> TO <devno range>
1869 VARY OFF PATH <PATH> FROM <devno range>
1870 This is used to switch on or off channel paths to devices.
1872 Q CHPID <channel path ID>
1873 This displays state of devices using this channel path
1874 D SCHIB <subchannel>
1875 This displays the subchannel information SCHIB block for the device.
1876 this I believe is also only available to administrators.
1878 defines a virtual CTC channel to channel connection
1879 2 need to be defined on each guest for the CTC driver to use.
1880 COUPLE devno userid remote devno
1881 Joins a local virtual device to a remote virtual device
1882 ( commonly used for the CTC driver ).
1884 Building a VM ramdisk under CMS which linux can use
1885 def vfb-<blocksize> <subchannel> <number blocks>
1886 blocksize is commonly 4096 for linux.
1888 format <subchannel> <driver letter e.g. x> (blksize <blocksize>
1890 Sharing a disk between multiple guests
1891 LINK userid devno1 devno2 mode password
1897 N.B. if compiling for debugging gdb works better without optimisation
1898 ( see Compiling programs for debugging )
1902 gdb <victim program> <optional corefile>
1906 help: gives help on commands
1910 Note gdb's online help is very good use it.
1915 info registers: displays registers other than floating point.
1916 info all-registers: displays floating points as well.
1917 disassemble: disassembles
1919 disassemble without parameters will disassemble the current function
1920 disassemble $pc $pc+10
1922 Viewing & modifying variables
1923 -----------------------------
1924 print or p: displays variable or register
1925 e.g. p/x $sp will display the stack pointer
1927 display: prints variable or register each time program stops
1929 display/x $pc will display the program counter
1932 undisplay : undo's display's
1934 info breakpoints: shows all current breakpoints
1936 info stack: shows stack back trace ( if this doesn't work too well, I'll show you the
1937 stacktrace by hand below ).
1939 info locals: displays local variables.
1941 info args: display current procedure arguments.
1943 set args: will set argc & argv each time the victim program is invoked.
1945 set <variable>=value
1953 step: steps n lines of sourcecode
1955 step 100 steps 100 lines of code.
1957 next: like step except this will not step into subroutines
1959 stepi: steps a single machine code instruction.
1962 nexti: steps a single machine code instruction but will not step into subroutines.
1964 finish: will run until exit of the current routine
1966 run: (re)starts a program
1968 cont: continues a program
1986 Here's a really useful one for large programs
1988 Set a breakpoint for all functions matching REGEXP
1991 will set a breakpoint with all functions with 390 in their name.
1994 lists all breakpoints
1996 delete: delete breakpoint by number or delete them all
1998 delete 1 will delete the first breakpoint
1999 delete will delete them all
2001 watch: This will set a watchpoint ( usually hardware assisted ),
2002 This will watch a variable till it changes
2004 watch cnt, will watch the variable cnt till it changes.
2005 As an aside unfortunately gdb's, architecture independent watchpoint code
2006 is inconsistent & not very good, watchpoints usually work but not always.
2008 info watchpoints: Display currently active watchpoints
2010 condition: ( another useful one )
2011 Specify breakpoint number N to break only if COND is true.
2012 Usage is `condition N COND', where N is an integer and COND is an
2013 expression to be evaluated whenever breakpoint N is reached.
2017 User defined functions/macros
2018 -----------------------------
2019 define: ( Note this is very very useful,simple & powerful )
2020 usage define <name> <list of commands> end
2022 examples which you should consider putting into .gdbinit in your home directory
2025 disassemble $pc $pc+10
2030 disassemble $pc $pc+10
2034 Other hard to classify stuff
2035 ----------------------------
2037 sends the victim program a signal.
2038 e.g. signal 3 will send a SIGQUIT.
2041 what gdb does when the victim receives certain signals.
2045 list lists current function source
2046 list 1,10 list first 10 lines of current file.
2051 Adds directories to be searched for source if gdb cannot find the source.
2052 (note it is a bit sensitive about slashes)
2053 e.g. To add the root of the filesystem to the searchpath do
2058 This calls a function in the victim program, this is pretty powerful
2060 (gdb) call printf("hello world")
2064 You might now be thinking that the line above didn't work, something extra had to be done.
2065 (gdb) call fflush(stdout)
2067 As an aside the debugger also calls malloc & free under the hood
2068 to make space for the "hello world" string.
2074 1) command completion works just like bash
2075 ( if you are a bad typist like me this really helps )
2076 e.g. hit br <TAB> & cursor up & down :-).
2078 2) if you have a debugging problem that takes a few steps to recreate
2079 put the steps into a file called .gdbinit in your current working directory
2080 if you have defined a few extra useful user defined commands put these in
2081 your home directory & they will be read each time gdb is launched.
2083 A typical .gdbinit file might be.
2086 break runtime_exception
2090 stack chaining in gdb by hand
2091 -----------------------------
2092 This is done using a the same trick described for VM
2093 p/x (*($sp+56))&0x7fffffff get the first backchain.
2096 Replace 56 with 112 & ignore the &0x7fffffff
2097 in the macros below & do nasty casts to longs like the following
2098 as gdb unfortunately deals with printed arguments as ints which
2099 messes up everything.
2100 i.e. here is a 3rd backchain dereference
2101 p/x *(long *)(***(long ***)$sp+112)
2108 info symbol (*($sp+56))&0x7fffffff
2109 you might see something like.
2110 rl_getc + 36 in section .text telling you what is located at address 0x528f18
2112 p/x (*(*$sp+56))&0x7fffffff
2116 info symbol (*(*$sp+56))&0x7fffffff
2117 rl_read_key + 180 in section .text
2119 p/x (*(**$sp+56))&0x7fffffff
2122 Disassembling instructions without debug info
2123 ---------------------------------------------
2124 gdb typically complains if there is a lack of debugging
2125 symbols in the disassemble command with
2126 "No function contains specified address." To get around
2128 x/<number lines to disassemble>xi <address>
2134 Note: Remember gdb has history just like bash you don't need to retype the
2135 whole line just use the up & down arrows.
2141 From your linuxbox do
2142 man gdb or info gdb.
2147 A core dump is a file generated by the kernel ( if allowed ) which contains the registers,
2148 & all active pages of the program which has crashed.
2149 From this file gdb will allow you to look at the registers & stack trace & memory of the
2150 program as if it just crashed on your system, it is usually called core & created in the
2151 current working directory.
2152 This is very useful in that a customer can mail a core dump to a technical support department
2153 & the technical support department can reconstruct what happened.
2154 Provided they have an identical copy of this program with debugging symbols compiled in &
2155 the source base of this build is available.
2156 In short it is far more useful than something like a crash log could ever hope to be.
2158 In theory all that is missing to restart a core dumped program is a kernel patch which
2159 will do the following.
2160 1) Make a new kernel task structure
2161 2) Reload all the dumped pages back into the kernel's memory management structures.
2162 3) Do the required clock fixups
2163 4) Get all files & network connections for the process back into an identical state ( really difficult ).
2164 5) A few more difficult things I haven't thought of.
2168 Why have I never seen one ?.
2169 Probably because you haven't used the command
2170 ulimit -c unlimited in bash
2171 to allow core dumps, now do
2173 to verify that the limit was accepted.
2176 To create this I'm going to do
2179 to launch gdb (my victim app. ) now be bad & do the following from another
2180 telnet/xterm session to the same machine
2182 kill -SIGSEGV <gdb's pid>
2183 or alternatively use killall -SIGSEGV gdb if you have the killall command.
2184 Now look at the core dump.
2186 Displays the following
2188 Copyright 1998 Free Software Foundation, Inc.
2189 GDB is free software, covered by the GNU General Public License, and you are
2190 welcome to change it and/or distribute copies of it under certain conditions.
2191 Type "show copying" to see the conditions.
2192 There is absolutely no warranty for GDB. Type "show warranty" for details.
2193 This GDB was configured as "s390-ibm-linux"...
2194 Core was generated by `./gdb'.
2195 Program terminated with signal 11, Segmentation fault.
2196 Reading symbols from /usr/lib/libncurses.so.4...done.
2197 Reading symbols from /lib/libm.so.6...done.
2198 Reading symbols from /lib/libc.so.6...done.
2199 Reading symbols from /lib/ld-linux.so.2...done.
2200 #0 0x40126d1a in read () from /lib/libc.so.6
2201 Setting up the environment for debugging gdb.
2202 Breakpoint 1 at 0x4dc6f8: file utils.c, line 471.
2203 Breakpoint 2 at 0x4d87a4: file top.c, line 2609.
2204 (top-gdb) info stack
2205 #0 0x40126d1a in read () from /lib/libc.so.6
2206 #1 0x528f26 in rl_getc (stream=0x7ffffde8) at input.c:402
2207 #2 0x528ed0 in rl_read_key () at input.c:381
2208 #3 0x5167e6 in readline_internal_char () at readline.c:454
2209 #4 0x5168ee in readline_internal_charloop () at readline.c:507
2210 #5 0x51692c in readline_internal () at readline.c:521
2211 #6 0x5164fe in readline (prompt=0x7ffff810 "\177
\81ÿ
\81øx\177
\81ÿ
\81÷
\81Ø\177
\81ÿ
\81øx
\81À")
2213 #7 0x4d7a8a in command_line_input (prompt=0x564420 "(gdb) ", repeat=1,
2214 annotation_suffix=0x4d6b44 "prompt") at top.c:2091
2215 #8 0x4d6cf0 in command_loop () at top.c:1345
2216 #9 0x4e25bc in main (argc=1, argv=0x7ffffdf4) at main.c:635
2221 This is a program which lists the shared libraries which a library needs,
2222 Note you also get the relocations of the shared library text segments which
2223 help when using objdump --source.
2227 libncurses.so.4 => /usr/lib/libncurses.so.4 (0x40018000)
2228 libm.so.6 => /lib/libm.so.6 (0x4005e000)
2229 libc.so.6 => /lib/libc.so.6 (0x40084000)
2230 /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)
2233 Debugging shared libraries
2234 ==========================
2235 Most programs use shared libraries, however it can be very painful
2236 when you single step instruction into a function like printf for the
2237 first time & you end up in functions like _dl_runtime_resolve this is
2238 the ld.so doing lazy binding, lazy binding is a concept in ELF where
2239 shared library functions are not loaded into memory unless they are
2240 actually used, great for saving memory but a pain to debug.
2241 To get around this either relink the program -static or exit gdb type
2242 export LD_BIND_NOW=true this will stop lazy binding & restart the gdb'ing
2243 the program in question.
2249 As modules are dynamically loaded into the kernel their address can be
2250 anywhere to get around this use the -m option with insmod to emit a load
2251 map which can be piped into a file if required.
2253 The proc file system
2254 ====================
2256 It is a filesystem created by the kernel with files which are created on demand
2257 by the kernel if read, or can be used to modify kernel parameters,
2258 it is a powerful concept.
2262 cat /proc/sys/net/ipv4/ip_forward
2263 On my machine outputs
2265 telling me ip_forwarding is not on to switch it on I can do
2266 echo 1 > /proc/sys/net/ipv4/ip_forward
2268 cat /proc/sys/net/ipv4/ip_forward
2269 On my machine now outputs
2271 IP forwarding is on.
2272 There is a lot of useful info in here best found by going in & having a look around,
2273 so I'll take you through some entries I consider important.
2275 All the processes running on the machine have their own entry defined by
2277 So lets have a look at the init process
2285 This contains numerical entries of all the open files,
2286 some of these you can cat e.g. stdout (2)
2291 00400000-00478000 r-xp 00000000 5f:00 4103 /bin/bash
2292 00478000-0047e000 rw-p 00077000 5f:00 4103 /bin/bash
2293 0047e000-00492000 rwxp 00000000 00:00 0
2294 40000000-40015000 r-xp 00000000 5f:00 14382 /lib/ld-2.1.2.so
2295 40015000-40016000 rw-p 00014000 5f:00 14382 /lib/ld-2.1.2.so
2296 40016000-40017000 rwxp 00000000 00:00 0
2297 40017000-40018000 rw-p 00000000 00:00 0
2298 40018000-4001b000 r-xp 00000000 5f:00 14435 /lib/libtermcap.so.2.0.8
2299 4001b000-4001c000 rw-p 00002000 5f:00 14435 /lib/libtermcap.so.2.0.8
2300 4001c000-4010d000 r-xp 00000000 5f:00 14387 /lib/libc-2.1.2.so
2301 4010d000-40111000 rw-p 000f0000 5f:00 14387 /lib/libc-2.1.2.so
2302 40111000-40114000 rw-p 00000000 00:00 0
2303 40114000-4011e000 r-xp 00000000 5f:00 14408 /lib/libnss_files-2.1.2.so
2304 4011e000-4011f000 rw-p 00009000 5f:00 14408 /lib/libnss_files-2.1.2.so
2305 7fffd000-80000000 rwxp ffffe000 00:00 0
2308 Showing us the shared libraries init uses where they are in memory
2309 & memory access permissions for each virtual memory area.
2311 /proc/1/cwd is a softlink to the current working directory.
2312 /proc/1/root is the root of the filesystem for this process.
2314 /proc/1/mem is the current running processes memory which you
2315 can read & write to like a file.
2316 strace uses this sometimes as it is a bit faster than the
2317 rather inefficient ptrace interface for peeking at DATA.
2336 SigPnd: 0000000000000000
2337 SigBlk: 0000000000000000
2338 SigIgn: 7fffffffd7f0d8fc
2339 SigCgt: 00000000280b2603
2340 CapInh: 00000000fffffeff
2341 CapPrm: 00000000ffffffff
2342 CapEff: 00000000fffffeff
2344 User PSW: 070de000 80414146
2345 task: 004b6000 tss: 004b62d8 ksp: 004b7ca8 pt_regs: 004b7f68
2347 00000400 00000000 0000000b 7ffffa90
2348 00000000 00000000 00000000 0045d9f4
2349 0045cafc 7ffffa90 7fffff18 0045cb08
2350 00010400 804039e8 80403af8 7ffff8b0
2352 00000000 00000000 00000000 00000000
2353 00000001 00000000 00000000 00000000
2354 00000000 00000000 00000000 00000000
2355 00000000 00000000 00000000 00000000
2356 Kernel BackChain CallChain BackChain CallChain
2357 004b7ca8 8002bd0c 004b7d18 8002b92c
2358 004b7db8 8005cd50 004b7e38 8005d12a
2360 Showing among other things memory usage & status of some signals &
2361 the processes'es registers from the kernel task_structure
2362 as well as a backchain which may be useful if a process crashes
2363 in the kernel for some unknown reason.
2365 Some driver debugging techniques
2366 ================================
2369 Some of our drivers now support a "debug feature" in
2370 /proc/s390dbf see s390dbf.txt in the linux/Documentation directory
2373 to switch on the lcs "debug feature"
2374 echo 5 > /proc/s390dbf/lcs/level
2375 & then after the error occurred.
2376 cat /proc/s390dbf/lcs/sprintf >/logfile
2377 the logfile now contains some information which may help
2378 tech support resolve a problem in the field.
2382 high level debugging network drivers
2383 ------------------------------------
2384 ifconfig is a quite useful command
2385 it gives the current state of network drivers.
2387 If you suspect your network device driver is dead
2388 one way to check is type
2389 ifconfig <network device>
2391 You should see something like
2392 tr0 Link encap:16/4 Mbps Token Ring (New) HWaddr 00:04:AC:20:8E:48
2393 inet addr:9.164.185.132 Bcast:9.164.191.255 Mask:255.255.224.0
2394 UP BROADCAST RUNNING MULTICAST MTU:2000 Metric:1
2395 RX packets:246134 errors:0 dropped:0 overruns:0 frame:0
2396 TX packets:5 errors:0 dropped:0 overruns:0 carrier:0
2397 collisions:0 txqueuelen:100
2399 if the device doesn't say up
2401 /etc/rc.d/init.d/network start
2402 ( this starts the network stack & hopefully calls ifconfig tr0 up ).
2403 ifconfig looks at the output of /proc/net/dev & presents it in a more presentable form
2404 Now ping the device from a machine in the same subnet.
2405 if the RX packets count & TX packets counts don't increment you probably
2409 Do you see any hardware addresses in the cache if not you may have problems.
2411 ping -c 5 <broadcast_addr> i.e. the Bcast field above in the output of
2412 ifconfig. Do you see any replies from machines other than the local machine
2413 if not you may have problems. also if the TX packets count in ifconfig
2414 hasn't incremented either you have serious problems in your driver
2415 (e.g. the txbusy field of the network device being stuck on )
2416 or you may have multiple network devices connected.
2421 There is a new device layer for channel devices, some
2422 drivers e.g. lcs are registered with this layer.
2423 If the device uses the channel device layer you'll be
2424 able to find what interrupts it uses & the current state
2426 See the manpage chandev.8 &type cat /proc/chandev for more info.
2430 Starting points for debugging scripting languages etc.
2431 ======================================================
2435 bash -x <scriptname>
2436 e.g. bash -x /usr/bin/bashbug
2437 displays the following lines as it executes them.
2441 + CFLAGS= -DPROGRAM='bash' -DHOSTTYPE='i586' -DOSTYPE='linux-gnu' -DMACHTYPE='i586-pc-linux-gnu' -DSHELL -DHAVE_CONFIG_H -I. -I. -I./lib -O2 -pipe
2445 + MACHTYPE=i586-pc-linux-gnu
2447 perl -d <scriptname> runs the perlscript in a fully interactive debugger
2449 Type 'h' in the debugger for help.
2451 for debugging java type
2452 jdb <filename> another fully interactive gdb style debugger.
2453 & type ? in the debugger for help.
2459 This is now supported by linux for s/390 & z/Architecture.
2460 To enable it do compile the kernel with
2461 Kernel Hacking -> Magic SysRq Key Enabled
2462 echo "1" > /proc/sys/kernel/sysrq
2464 echo "8" >/proc/sys/kernel/printk
2465 To make printk output go to console.
2466 On 390 all commands are prefixed with
2469 ^-t will show tasks.
2470 ^-? or some unknown command will display help.
2471 The sysrq key reading is very picky ( I have to type the keys in an
2472 xterm session & paste them into the x3270 console )
2473 & it may be wise to predefine the keys as described in the VM hints above
2475 This is particularly useful for syncing disks unmounting & rebooting
2476 if the machine gets partially hung.
2478 Read Documentation/sysrq.txt for more info
2482 Enterprise Systems Architecture Reference Summary
2483 Enterprise Systems Architecture Principles of Operation
2484 Hartmut Penners s390 stack frame sheet.
2485 IBM Mainframe Channel Attachment a technology brief from a CISCO webpage
2486 Various bits of man & info pages of Linux.
2488 Various info & man pages.
2489 CMS Help on tracing commands.
2490 Linux for s/390 Elf Application Binary Interface
2491 Linux for z/Series Elf Application Binary Interface ( Both Highly Recommended )
2492 z/Architecture Principles of Operation SA22-7832-00
2493 Enterprise Systems Architecture/390 Reference Summary SA22-7209-01 & the
2494 Enterprise Systems Architecture/390 Principles of Operation SA22-7201-05
2498 Special thanks to Neale Ferguson who maintains a much
2499 prettier HTML version of this page at
2500 http://linuxvm.org/penguinvm/
2501 Bob Grainger Stefan Bader & others for reporting bugs