linux/Documentation/x86/kernel-stacks.rst
<<
>>
Prefs
   1.. SPDX-License-Identifier: GPL-2.0
   2
   3=============
   4Kernel Stacks
   5=============
   6
   7Kernel stacks on x86-64 bit
   8===========================
   9
  10Most of the text from Keith Owens, hacked by AK
  11
  12x86_64 page size (PAGE_SIZE) is 4K.
  13
  14Like all other architectures, x86_64 has a kernel stack for every
  15active thread.  These thread stacks are THREAD_SIZE (2*PAGE_SIZE) big.
  16These stacks contain useful data as long as a thread is alive or a
  17zombie. While the thread is in user space the kernel stack is empty
  18except for the thread_info structure at the bottom.
  19
  20In addition to the per thread stacks, there are specialized stacks
  21associated with each CPU.  These stacks are only used while the kernel
  22is in control on that CPU; when a CPU returns to user space the
  23specialized stacks contain no useful data.  The main CPU stacks are:
  24
  25* Interrupt stack.  IRQ_STACK_SIZE
  26
  27  Used for external hardware interrupts.  If this is the first external
  28  hardware interrupt (i.e. not a nested hardware interrupt) then the
  29  kernel switches from the current task to the interrupt stack.  Like
  30  the split thread and interrupt stacks on i386, this gives more room
  31  for kernel interrupt processing without having to increase the size
  32  of every per thread stack.
  33
  34  The interrupt stack is also used when processing a softirq.
  35
  36Switching to the kernel interrupt stack is done by software based on a
  37per CPU interrupt nest counter. This is needed because x86-64 "IST"
  38hardware stacks cannot nest without races.
  39
  40x86_64 also has a feature which is not available on i386, the ability
  41to automatically switch to a new stack for designated events such as
  42double fault or NMI, which makes it easier to handle these unusual
  43events on x86_64.  This feature is called the Interrupt Stack Table
  44(IST).  There can be up to 7 IST entries per CPU. The IST code is an
  45index into the Task State Segment (TSS). The IST entries in the TSS
  46point to dedicated stacks; each stack can be a different size.
  47
  48An IST is selected by a non-zero value in the IST field of an
  49interrupt-gate descriptor.  When an interrupt occurs and the hardware
  50loads such a descriptor, the hardware automatically sets the new stack
  51pointer based on the IST value, then invokes the interrupt handler.  If
  52the interrupt came from user mode, then the interrupt handler prologue
  53will switch back to the per-thread stack.  If software wants to allow
  54nested IST interrupts then the handler must adjust the IST values on
  55entry to and exit from the interrupt handler.  (This is occasionally
  56done, e.g. for debug exceptions.)
  57
  58Events with different IST codes (i.e. with different stacks) can be
  59nested.  For example, a debug interrupt can safely be interrupted by an
  60NMI.  arch/x86_64/kernel/entry.S::paranoidentry adjusts the stack
  61pointers on entry to and exit from all IST events, in theory allowing
  62IST events with the same code to be nested.  However in most cases, the
  63stack size allocated to an IST assumes no nesting for the same code.
  64If that assumption is ever broken then the stacks will become corrupt.
  65
  66The currently assigned IST stacks are:
  67
  68* ESTACK_DF.  EXCEPTION_STKSZ (PAGE_SIZE).
  69
  70  Used for interrupt 8 - Double Fault Exception (#DF).
  71
  72  Invoked when handling one exception causes another exception. Happens
  73  when the kernel is very confused (e.g. kernel stack pointer corrupt).
  74  Using a separate stack allows the kernel to recover from it well enough
  75  in many cases to still output an oops.
  76
  77* ESTACK_NMI.  EXCEPTION_STKSZ (PAGE_SIZE).
  78
  79  Used for non-maskable interrupts (NMI).
  80
  81  NMI can be delivered at any time, including when the kernel is in the
  82  middle of switching stacks.  Using IST for NMI events avoids making
  83  assumptions about the previous state of the kernel stack.
  84
  85* ESTACK_DB.  EXCEPTION_STKSZ (PAGE_SIZE).
  86
  87  Used for hardware debug interrupts (interrupt 1) and for software
  88  debug interrupts (INT3).
  89
  90  When debugging a kernel, debug interrupts (both hardware and
  91  software) can occur at any time.  Using IST for these interrupts
  92  avoids making assumptions about the previous state of the kernel
  93  stack.
  94
  95  To handle nested #DB correctly there exist two instances of DB stacks. On
  96  #DB entry the IST stackpointer for #DB is switched to the second instance
  97  so a nested #DB starts from a clean stack. The nested #DB switches
  98  the IST stackpointer to a guard hole to catch triple nesting.
  99
 100* ESTACK_MCE.  EXCEPTION_STKSZ (PAGE_SIZE).
 101
 102  Used for interrupt 18 - Machine Check Exception (#MC).
 103
 104  MCE can be delivered at any time, including when the kernel is in the
 105  middle of switching stacks.  Using IST for MCE events avoids making
 106  assumptions about the previous state of the kernel stack.
 107
 108For more details see the Intel IA32 or AMD AMD64 architecture manuals.
 109
 110
 111Printing backtraces on x86
 112==========================
 113
 114The question about the '?' preceding function names in an x86 stacktrace
 115keeps popping up, here's an indepth explanation. It helps if the reader
 116stares at print_context_stack() and the whole machinery in and around
 117arch/x86/kernel/dumpstack.c.
 118
 119Adapted from Ingo's mail, Message-ID: <20150521101614.GA10889@gmail.com>:
 120
 121We always scan the full kernel stack for return addresses stored on
 122the kernel stack(s) [1]_, from stack top to stack bottom, and print out
 123anything that 'looks like' a kernel text address.
 124
 125If it fits into the frame pointer chain, we print it without a question
 126mark, knowing that it's part of the real backtrace.
 127
 128If the address does not fit into our expected frame pointer chain we
 129still print it, but we print a '?'. It can mean two things:
 130
 131 - either the address is not part of the call chain: it's just stale
 132   values on the kernel stack, from earlier function calls. This is
 133   the common case.
 134
 135 - or it is part of the call chain, but the frame pointer was not set
 136   up properly within the function, so we don't recognize it.
 137
 138This way we will always print out the real call chain (plus a few more
 139entries), regardless of whether the frame pointer was set up correctly
 140or not - but in most cases we'll get the call chain right as well. The
 141entries printed are strictly in stack order, so you can deduce more
 142information from that as well.
 143
 144The most important property of this method is that we _never_ lose
 145information: we always strive to print _all_ addresses on the stack(s)
 146that look like kernel text addresses, so if debug information is wrong,
 147we still print out the real call chain as well - just with more question
 148marks than ideal.
 149
 150.. [1] For things like IRQ and IST stacks, we also scan those stacks, in
 151       the right order, and try to cross from one stack into another
 152       reconstructing the call chain. This works most of the time.
 153