https://github.com/torvalds/linux
Revision 9230a0b65b47fe6856c4468ec0175c4987e5bede authored by Dave Chinner on 20 November 2018, 06:50:08 UTC, committed by Darrick J. Wong on 21 November 2018, 18:10:53 UTC
Long saga. There have been days spent following this through dead end
after dead end in multi-GB event traces. This morning, after writing
a trace-cmd wrapper that enabled me to be more selective about XFS
trace points, I discovered that I could get just enough essential
tracepoints enabled that there was a 50:50 chance the fsx config
would fail at ~115k ops. If it didn't fail at op 115547, I stopped
fsx at op 115548 anyway.

That gave me two traces - one where the problem manifested, and one
where it didn't. After refining the traces to have the necessary
information, I found that in the failing case there was a real
extent in the COW fork compared to an unwritten extent in the
working case.

Walking back through the two traces to the point where the CWO fork
extents actually diverged, I found that the bad case had an extra
unwritten extent in it. This is likely because the bug it led me to
had triggered multiple times in those 115k ops, leaving stray
COW extents around. What I saw was a COW delalloc conversion to an
unwritten extent (as they should always be through
xfs_iomap_write_allocate()) resulted in a /written extent/:

xfs_writepage:        dev 259:0 ino 0x83 pgoff 0x17000 size 0x79a00 offset 0 length 0
xfs_iext_remove:      dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/2 offset 32 block 152 count 20 flag 1 caller xfs_bmap_add_extent_delay_real
xfs_bmap_pre_update:  dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 4503599627239429 count 31 flag 0 caller xfs_bmap_add_extent_delay_real
xfs_bmap_post_update: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 121 count 51 flag 0 caller xfs_bmap_add_ex

Basically, Cow fork before:

	0 1            32          52
	+H+DDDDDDDDDDDD+UUUUUUUUUUU+
	   PREV		RIGHT

COW delalloc conversion allocates:

	  1	       32
	  +uuuuuuuuuuuu+
	  NEW

And the result according to the xfs_bmap_post_update trace was:

	0 1            32          52
	+H+wwwwwwwwwwwwwwwwwwwwwwww+
	   PREV

Which is clearly wrong - it should be a merged unwritten extent,
not an unwritten extent.

That lead me to look at the LEFT_FILLING|RIGHT_FILLING|RIGHT_CONTIG
case in xfs_bmap_add_extent_delay_real(), and sure enough, there's
the bug.

It takes the old delalloc extent (PREV) and adds the length of the
RIGHT extent to it, takes the start block from NEW, removes the
RIGHT extent and then updates PREV with the new extent.

What it fails to do is update PREV.br_state. For delalloc, this is
always XFS_EXT_NORM, while in this case we are converting the
delayed allocation to unwritten, so it needs to be updated to
XFS_EXT_UNWRITTEN. This LF|RF|RC case does not do this, and so
the resultant extent is always written.

And that's the bug I've been chasing for a week - a bmap btree bug,
not a reflink/dedupe/copy_file_range bug, but a BMBT bug introduced
with the recent in core extent tree scalability enhancements.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
1 parent 2c30717
Raw File
Tip revision: 9230a0b65b47fe6856c4468ec0175c4987e5bede authored by Dave Chinner on 20 November 2018, 06:50:08 UTC
xfs: delalloc -> unwritten COW fork allocation can go wrong
Tip revision: 9230a0b
kmod.c
/*
 * kmod - the kernel module loader
 */
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/sched/task.h>
#include <linux/binfmts.h>
#include <linux/syscalls.h>
#include <linux/unistd.h>
#include <linux/kmod.h>
#include <linux/slab.h>
#include <linux/completion.h>
#include <linux/cred.h>
#include <linux/file.h>
#include <linux/fdtable.h>
#include <linux/workqueue.h>
#include <linux/security.h>
#include <linux/mount.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/resource.h>
#include <linux/notifier.h>
#include <linux/suspend.h>
#include <linux/rwsem.h>
#include <linux/ptrace.h>
#include <linux/async.h>
#include <linux/uaccess.h>

#include <trace/events/module.h>

/*
 * Assuming:
 *
 * threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE,
 *		       (u64) THREAD_SIZE * 8UL);
 *
 * If you need less than 50 threads would mean we're dealing with systems
 * smaller than 3200 pages. This assuems you are capable of having ~13M memory,
 * and this would only be an be an upper limit, after which the OOM killer
 * would take effect. Systems like these are very unlikely if modules are
 * enabled.
 */
#define MAX_KMOD_CONCURRENT 50
static atomic_t kmod_concurrent_max = ATOMIC_INIT(MAX_KMOD_CONCURRENT);
static DECLARE_WAIT_QUEUE_HEAD(kmod_wq);

/*
 * This is a restriction on having *all* MAX_KMOD_CONCURRENT threads
 * running at the same time without returning. When this happens we
 * believe you've somehow ended up with a recursive module dependency
 * creating a loop.
 *
 * We have no option but to fail.
 *
 * Userspace should proactively try to detect and prevent these.
 */
#define MAX_KMOD_ALL_BUSY_TIMEOUT 5

/*
	modprobe_path is set via /proc/sys.
*/
char modprobe_path[KMOD_PATH_LEN] = "/sbin/modprobe";

static void free_modprobe_argv(struct subprocess_info *info)
{
	kfree(info->argv[3]); /* check call_modprobe() */
	kfree(info->argv);
}

static int call_modprobe(char *module_name, int wait)
{
	struct subprocess_info *info;
	static char *envp[] = {
		"HOME=/",
		"TERM=linux",
		"PATH=/sbin:/usr/sbin:/bin:/usr/bin",
		NULL
	};

	char **argv = kmalloc(sizeof(char *[5]), GFP_KERNEL);
	if (!argv)
		goto out;

	module_name = kstrdup(module_name, GFP_KERNEL);
	if (!module_name)
		goto free_argv;

	argv[0] = modprobe_path;
	argv[1] = "-q";
	argv[2] = "--";
	argv[3] = module_name;	/* check free_modprobe_argv() */
	argv[4] = NULL;

	info = call_usermodehelper_setup(modprobe_path, argv, envp, GFP_KERNEL,
					 NULL, free_modprobe_argv, NULL);
	if (!info)
		goto free_module_name;

	return call_usermodehelper_exec(info, wait | UMH_KILLABLE);

free_module_name:
	kfree(module_name);
free_argv:
	kfree(argv);
out:
	return -ENOMEM;
}

/**
 * __request_module - try to load a kernel module
 * @wait: wait (or not) for the operation to complete
 * @fmt: printf style format string for the name of the module
 * @...: arguments as specified in the format string
 *
 * Load a module using the user mode module loader. The function returns
 * zero on success or a negative errno code or positive exit code from
 * "modprobe" on failure. Note that a successful module load does not mean
 * the module did not then unload and exit on an error of its own. Callers
 * must check that the service they requested is now available not blindly
 * invoke it.
 *
 * If module auto-loading support is disabled then this function
 * becomes a no-operation.
 */
int __request_module(bool wait, const char *fmt, ...)
{
	va_list args;
	char module_name[MODULE_NAME_LEN];
	int ret;

	/*
	 * We don't allow synchronous module loading from async.  Module
	 * init may invoke async_synchronize_full() which will end up
	 * waiting for this task which already is waiting for the module
	 * loading to complete, leading to a deadlock.
	 */
	WARN_ON_ONCE(wait && current_is_async());

	if (!modprobe_path[0])
		return 0;

	va_start(args, fmt);
	ret = vsnprintf(module_name, MODULE_NAME_LEN, fmt, args);
	va_end(args);
	if (ret >= MODULE_NAME_LEN)
		return -ENAMETOOLONG;

	ret = security_kernel_module_request(module_name);
	if (ret)
		return ret;

	if (atomic_dec_if_positive(&kmod_concurrent_max) < 0) {
		pr_warn_ratelimited("request_module: kmod_concurrent_max (%u) close to 0 (max_modprobes: %u), for module %s, throttling...",
				    atomic_read(&kmod_concurrent_max),
				    MAX_KMOD_CONCURRENT, module_name);
		ret = wait_event_killable_timeout(kmod_wq,
						  atomic_dec_if_positive(&kmod_concurrent_max) >= 0,
						  MAX_KMOD_ALL_BUSY_TIMEOUT * HZ);
		if (!ret) {
			pr_warn_ratelimited("request_module: modprobe %s cannot be processed, kmod busy with %d threads for more than %d seconds now",
					    module_name, MAX_KMOD_CONCURRENT, MAX_KMOD_ALL_BUSY_TIMEOUT);
			return -ETIME;
		} else if (ret == -ERESTARTSYS) {
			pr_warn_ratelimited("request_module: sigkill sent for modprobe %s, giving up", module_name);
			return ret;
		}
	}

	trace_module_request(module_name, wait, _RET_IP_);

	ret = call_modprobe(module_name, wait ? UMH_WAIT_PROC : UMH_WAIT_EXEC);

	atomic_inc(&kmod_concurrent_max);
	wake_up(&kmod_wq);

	return ret;
}
EXPORT_SYMBOL(__request_module);
back to top