Difference between revisions of "Kernel Small Stacks"
(→Stack tracer in ftrace (or perf?))
(→Stack tracer in ftrace)
|Line 255:||Line 255:|
=== Stack tracer in ftrace ===
=== Stack tracer in ftrace ===
Revision as of 10:24, 21 March 2013
Here is some random information about small kernel stack sizes.
The default stack size for a process running in kernel space is 8K (as of 2011).
There used to be an option on x86 to reduce the stack size to 4K. And indeed there were efforts in 2006 to make this the default stack size. However, using a small stack opens up the dangerous possibility that the stack will overflow, causing a kernel hang.
Besides wasting memory, if the stack space is not really needed, 8K stacks also have an effect on, and are affected by, general kernel memory allocation. To create an 8K stack requires an order-1 allocation, meaning that 2 contiguous physical pages must be allocated together in order to create a new process stack. If memory has become fragmented, it may be impossible to fulfill an order-1 allocation, even though individual pages of physical memory may be free. Thus 4K stack allocations (order-0 allocations) are more likely to succeed. This is important for systems operating under extreme memory pressure.
There were years of debate on the kernel mailing list about whether 4K stacks should be the default, and lots of bug reports that ended up being caused stack overflows with 4K stacks enabled. The option to support 4k stacks on x86 was removed in 2010, from kernel version 2.6.37 with this commit: dcfa726280116dd31adad37da940f542663567d0
- 1 Historical information on 4K Stacks
- 2 Stack layout
- 3 Stack measuring/monitoring mechanisms
- 4 ARM 4K Stacks
- 5 Possible mixed stack size feature
- 6 Problems
Historical information on 4K Stacks
Here are some articles and links to information about 4K stacks:
- 4K Stacks in 2.6 - LWN.net, May 2004
- 4K stacks for everyone? - LWN.net, Sep 2005
- 4K stacks - again - LWN.net, Nov 2005
- 4K stacks by default? - LWN.net, Apr 2008
- http://www.redhat.com/archives/rhl-list/2007-March/msg00854.html - James Wilkinson message about memory fragmentation.
The kernel stack is laid out with the stack pointer at the top of each stack (at the highest stack address), growing downward for each function call and stack allocation. The thread_info structure for a process is at the bottom of the stack. There is no physical mechanism to detect, at allocation time, if the stack pointer wanders into the thread_info area of the stack. Hence, if the stack overflows (the stack pointer goes into the thread_info area), the behavior of the system is undefined.
top +----------------+ | return vals | | & local vars | | ... | | | | | | 0's | | thread_info | bottom +----------------+
Stack measuring/monitoring mechanisms
Because of previous efforts to conserve stack space, there are actually a few different mechanisms for monitoring the kernel stack usage. Some tools report on the static size of stack usage by kernel functions (a check which is done by either the compiler or a separate tool operating on the kernel binary), and some mechanisms can report on actual stack utilization at runtime.
This kernel configuration option passes an option to the compiler to cause it to emit a warning when a static stack size for a routine is detected that is larger than the specified threshold. It requires gcc version 4.4 or later in order to work.
The gcc option used is "-Wframe-larger-than=xxx".
By default, CONFIG_FRAME_WARN has the value of 1024, but you can set it to any value from 0 to 8192.
Here is a sample of build output with this option configured to 256:
... CC ipc/msg.o CC ipc/sem.o .../linux-3.0.y/ipc/sem.c: In function 'semctl_main.clone.7': .../linux-3.0.y/ipc/sem.c:1021:1: warning: the frame size of 520 bytes is larger than 256 bytes .../linux-3.0.y/ipc/sem.c: In function 'sys_semtimedop': .../linux-3.0.y/ipc/sem.c:1514:1: warning: the frame size of 472 bytes is larger than 256 bytes CC ipc/shm.o CC ipc/ipcns_notifier.o
The kernel source includes a script to perform static stack analysis called scripts/checkstack.pl.
Usage is as follows:
$(CROSS_COMPILE)objdump -d vmlinux | scripts/checkstack.pl [arch]
Replace [arch] with the architecture of the kernel being analyzed. Several architectures are supported, including arm, mips and x86. You should use a cross-objdump that matches the architecture you compiled the kernel for. For example, if you used: arm-gnueabi-linux-gcc as your compiler, you would use arm-gnueabi-linux-objdump as your object dump program. This should have been included in your cross-compiler toolchain package.
Below is some sample output from using checkstack.pl. Note that the file is first dumped to an assembly file (.S), and then piped to checkstack.pl. You can examine the assembly file to see in detail the instructions used to reserve space on the stack, for routines of interest found by checkstack.pl.
An item in brackets is a module name, in case of a loadable module. The number at end is stack depth detected for function. The Leading value is the address of the stack reservation code.
$ arm-eabi-objdummp -d vmlinux -o vmlinux-arm.S $ cat vmlinux-arm.S | scripts/checkstack.pl arm 0x0012c858 nlmclnt_reclaim [vmlinux-arm.o]: 720 0x0025748c do_tcp_getsockopt.clone.11 [vmlinux-arm.o]: 552 0x00258d04 do_tcp_setsockopt.clone.14 [vmlinux-arm.o]: 544 0x000b2db4 do_sys_poll [vmlinux-arm.o]: 532 0x00138744 semctl_main.clone.7 [vmlinux-arm.o]: 532 0x00138ec4 sys_semtimedop [vmlinux-arm.o]: 484 0x000c5618 default_file_splice_read [vmlinux-arm.o]: 436 0x00251de4 do_ip_setsockopt.clone.22 [vmlinux-arm.o]: 416 0x00191fd4 extract_buf [vmlinux-arm.o]: 408 0x0019bc24 loop_get_status_old [vmlinux-arm.o]: 396 0x000e6f88 do_task_stat [vmlinux-arm.o]: 380 0x0019b8f0 loop_set_status_old [vmlinux-arm.o]: 380 0x002078f0 snd_ctl_elem_add_user [vmlinux-arm.o]: 376 0x0026267c tcp_make_synack [vmlinux-arm.o]: 372 0x00127be4 nfs_dns_parse [vmlinux-arm.o]: 368 0x000b2240 do_select [vmlinux-arm.o]: 340 0x001f6f10 mmc_blk_issue_rw_rq [vmlinux-arm.o]: 340 0x001726a0 fb_set_var [vmlinux-arm.o]: 336 0x000c58d0 __generic_file_splice_read [vmlinux-arm.o]: 316 0x0022a074 dev_seq_printf_stats [vmlinux-arm.o]: 316 0x0006383c tracing_splice_read_pipe [vmlinux-arm.o]: 308 0x000c53c8 vmsplice_to_pipe [vmlinux-arm.o]: 308 0x002512b4 do_ip_getsockopt [vmlinux-arm.o]: 304 0x00225f68 skb_splice_bits [vmlinux-arm.o]: 300
Below are some results for static analysis of function stack depth in the Linux kernel, using 'stack_size'. (stack_size is a custom tool written by Tim Bird, before he found out about checkstack.pl.)
See this kernel message for a patch containing 'stack_size': https://lkml.org/lkml/2011/10/18/479
The following results include the reduction in size for 'struct poll_wqueue':
$ ./stack_size vmlinux-arm ============ RESULTS =============== number of functions = 14371 max function stack depth= 736 function with max depth = nlmclnt_reclaim Function Name Stack Depth ===================== =========== __generic_file_splice_read 352 do_select 376 loop_set_status_old 392 snd_ctl_elem_add_user 408 extract_buf 432 default_file_splice_read 472 sys_semtimedop 520 semctl_main.clone.7 560 do_sys_poll 568 nlmclnt_reclaim 736
$ ./show_stacks_x86_64.py vmlinux-x86_64.o ============ RESULTS =============== number of functions = 29587 max function stack depth= 1208 function with max depth = security_load_policy Function Name Stack Depth ===================== =========== x86_schedule_events 632 drm_crtc_helper_set_mode 632 sys_semtimedop 664 do_task_stat 712 node_read_meminfo 760 default_file_splice_read 792 do_select 920 nlmclnt_reclaim 936 do_sys_poll 1048 security_load_policy 1208
There is kernel feature to output the stack usage of each process, as well as the process that uses the most stack in the system. This is controlled by the kernel configuration option CONFIG_DEBUG_STACK_USAGE.
This option modifies the process creation path so that the stack is initialized with all zeros. At any time, a request can be made to measure the stack depth of all running processes. This is calculated by measuring the amount of zeros from the end of thread_info to the first non-zero item on each stack.
In more detail, it does the following:
- at process creation time, fills the stack with zeros (kernel/fork.c)
- on sysrq 't', show free space, from call to stack_not_used() (kernel/sched.c)
- it shows as 0 otherwise ??
- define check_stack_usage(), which emits printks on each low-water hit
- low-water appears to be global over all stacks
- check_stack_usage() is only called on process exit, so you might not know about a problem process until very late
- stack_not_used() is defined in include/linux/sched.h. It counts the number of zero bytes following the end of thread_info going up.
As the systems runs, any time the stack low-water mark is exceeded, then the kernel prints a report (logs it to the kernel message log). This can be viewed with the dmesg command:
Here is example output, greping the kernel message log for "greatest":
# dmesg | grep greatest kworker/u:0 used greatest stack depth: 10564 bytes left busybox used greatest stack depth: 9512 bytes left busybox used greatest stack depth: 9504 bytes left grep used greatest stack depth: 9372 bytes left init used greatest stack depth: 9028 bytes left
To get a report on the stack usage of currently running processes, you use 't' with sysrq. For example:
$ echo t >/proc/sysrq-trigger
A stack dump for each process is shown, along with stack usage information.
Here is some sample output:
$ echo t >/proc/sysrq-trigger $ dmesg | grep -v [ task PC stack pid father init S 802af8b0 932 1 0 0x00000000 kthreadd S 802af8b0 2496 2 0 0x00000000 ksoftirqd/0 S 802af8b0 2840 3 2 0x00000000 kworker/0:0 S 802af8b0 2776 4 2 0x00000000 kworker/u:0 S 802af8b0 2548 5 2 0x00000000 migration/0 S 802af8b0 2704 6 2 0x00000000 migration/1 S 802af8b0 2704 7 2 0x00000000 kworker/1:0 S 802af8b0 2560 8 2 0x00000000 ksoftirqd/1 S 802af8b0 3024 9 2 0x00000000 khelper S 802af8b0 2824 10 2 0x00000000 sync_supers S 802af8b0 2872 11 2 0x00000000 bdi-default S 802af8b0 2584 12 2 0x00000000 kblockd S 802af8b0 2824 13 2 0x00000000 khubd S 802af8b0 2744 14 2 0x00000000 rpciod S 802af8b0 3024 15 2 0x00000000 kworker/0:1 S 802af8b0 1240 16 2 0x00000000 kswapd0 S 802af8b0 2848 17 2 0x00000000 fsnotify_mark S 802af8b0 2632 18 2 0x00000000 nfsiod S 802af8b0 3024 19 2 0x00000000 kworker/u:1 S 802af8b0 2840 20 2 0x00000000 hoge S 802af8b0 3024 23 2 0x00000000 kworker/1:1 S 802af8b0 1716 24 2 0x00000000 flush-0:13 S 802af8b0 2528 28 2 0x00000000 telnetd S 802af8b0 1848 48 1 0x00000000 ash R running 1264 56 1 0x00000000
Stack tracer in ftrace
For detailed instructions, see: http://elinux.org/Ftrace#find_deepest_kernel_stack
- Turn on CONFIG_STACK_TRACER in kernel config
- pass 'stacktrace' on kernel command line, or at runtime do:
- echo 1 >/proc/sys/kernel/stack_tracer_enabled
- mount -t debugfs none /sys/kernel/debug
- <do some stuff>
- cat /sys/kernel/debug/tracing/stack_trace
Stack limit patches
Sony has a series of patches which implement a stack guard page, and use that to show a backtrace if the process uses more than a specified amount in its kernel stack. In essence, this creates a hard failure for a controlled stack overflow event.
These patches do the following: Add the config options:
- CONFIG_SNSC_DEBUG_STACK_LIMIT - perform stack layout changes
- CONFIG_SNSC_SUPPORT_4KB_MAPPING - re-map kernel memory for 4K TLB mappings
- CONFIG_SNSC_DEBUG_STACK_LIMIT_MANUAL - allow a user-specified starting stack size
the patches do several things:
- change the stack layout to place thread_info at the top of the stack rather than at the bottom
- change the stack size to 16KB (order 2)
- allow configuring the default starting position of the stack, to simulate an arbitrary stack size (default is right below thread_info)
- remap the kernel memory so that 4KB mappings are used
- this allows unmapping the bottom page of the stack, so that a fault occurs when the page is accessed (on a stack overflow for a manually configured small stack)
- this is only turned on if you specify 'use_4kb_mapping' on the
- unmap the bottom page of the stack (the guard page)
- this is only turned on if you specify 'unmap_stack'
[ FIXTHIS - need to add more to this section]
ARM 4K Stacks
In October of 2011, Tim Bird submitted patches to add 4K stack support for the ARM architecture to the Linux kernel. The patches he submitted are here:
- https://lkml.org/lkml/2011/10/18/476 - ARM 4Kstacks: introduction
- https://lkml.org/lkml/2011/10/18/477 - ARM 4Kstacks: Add support for 4K kernel stacks to ARM
- https://lkml.org/lkml/2011/10/18/479 - ARM: Add static kernel function stack size analyzer, for ARM
- https://lkml.org/lkml/2011/10/18/481 - ARM 4Kstacks: Decrease poll and select stack usage, when using 4K stacks
After some discussion, these patches were not accepted into mainline.
The following points were problems that needed to be addressed for this patch set:
* Should make this depend on CONFIG_EXPERT (to warn developers who attempt to use this) * Should add interrupt stacks to ARM to take pressure off of 4K stacks * Should determine if 4K stacks use case will cause ripple effect and lots of ifdefs and hard maintenance issues throughout the kernel. In particular, need to look at: * %pV recursion in printk. This is used by several file systems * question: for operation or just reporting??
Dave Chinner (here) wrote:
There's a good reason 4k stacks went away: it's simply not enough space for the deep 60+ function call stacks we see with even trivial storage stack configurations. The stack usage on 32 bit ARM and x86 is going to be similar, so you're going to be fighting a losing battle - there is no stack space that can be trimmed from most paths. To make matter worse, there's been stuff done to the storage stack that significantly increases stack usage since 4k stacks went away (e.g. the on-stack block plugging changes). And FWIW, XFS is widely used on ARM based NAS devices, so this isn't a theoretical problem I'm making up here...
This is a pretty good example of people denying a use case with a red herring.
Possible mixed stack size feature
One option for realizing most of the benefits of 4K stacks, while preserving more robustness, would be to utilize mixed stack sizes in the kernel.
Processes known to exercise only certain, stack-conservative, code paths in the kernel could utilize 4K stacks, and other processes could utilize 8K (or larger) stacks for safety purposes.
There would have to be a mechanism to support selecting the stack size at process creation time. One simple mechanism would be to introduce a child_stack_size parameter in thread_info, settable via /proc, and use this on the clone system call.
A system to support different-sized stacks by changing the stack size of already running processes would likely be too complicated to be practical.
Currently, the method of accessing the thread_info structure for a task in the kernel relies on the stack size of all processes being consistent among all processes (and being a power of two). A pointer to thread_info is obtained by masking the current stack pointer with a value dependent on the (global) size of the stack. With mixed stack sizes, a different mechanism would be needed to convert from stack pointer to thread_info. One method which might work would be to pre-allocate a stack pool for non-standard-sized stacks, and use pointer comparison to see if SP fell within the pool. If the pool was exhausted, the default stack size would be used.
This would work best in the case of a system with an identifiable number of processes which would use special-sized stacks.
This area has random notes for stack depth management issues:
The structure 'struct poll_wqueue' is a large data structure used for the select() and poll() system calls to manage a sub-set of the file descriptors being polled. This structure includes an array of wait queues which can be used immediately (without requiring or waiting for a memory allocation) for polling file I/O.
The number of entries in the array of wait queues can be controlled via macros in include/linux/poll.h
network lock manager for network filesystems. Not applicable to most embedded products (except possibly during development).
An selinux routine, not applicable to embedded.