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
random32.c
// SPDX-License-Identifier: GPL-2.0
/*
 * This is a maximally equidistributed combined Tausworthe generator
 * based on code from GNU Scientific Library 1.5 (30 Jun 2004)
 *
 * lfsr113 version:
 *
 * x_n = (s1_n ^ s2_n ^ s3_n ^ s4_n)
 *
 * s1_{n+1} = (((s1_n & 4294967294) << 18) ^ (((s1_n <<  6) ^ s1_n) >> 13))
 * s2_{n+1} = (((s2_n & 4294967288) <<  2) ^ (((s2_n <<  2) ^ s2_n) >> 27))
 * s3_{n+1} = (((s3_n & 4294967280) <<  7) ^ (((s3_n << 13) ^ s3_n) >> 21))
 * s4_{n+1} = (((s4_n & 4294967168) << 13) ^ (((s4_n <<  3) ^ s4_n) >> 12))
 *
 * The period of this generator is about 2^113 (see erratum paper).
 *
 * From: P. L'Ecuyer, "Maximally Equidistributed Combined Tausworthe
 * Generators", Mathematics of Computation, 65, 213 (1996), 203--213:
 * http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme.ps
 * ftp://ftp.iro.umontreal.ca/pub/simulation/lecuyer/papers/tausme.ps
 *
 * There is an erratum in the paper "Tables of Maximally Equidistributed
 * Combined LFSR Generators", Mathematics of Computation, 68, 225 (1999),
 * 261--269: http://www.iro.umontreal.ca/~lecuyer/myftp/papers/tausme2.ps
 *
 *      ... the k_j most significant bits of z_j must be non-zero,
 *      for each j. (Note: this restriction also applies to the
 *      computer code given in [4], but was mistakenly not mentioned
 *      in that paper.)
 *
 * This affects the seeding procedure by imposing the requirement
 * s1 > 1, s2 > 7, s3 > 15, s4 > 127.
 */

#include <linux/types.h>
#include <linux/percpu.h>
#include <linux/export.h>
#include <linux/jiffies.h>
#include <linux/random.h>
#include <linux/sched.h>
#include <asm/unaligned.h>

#ifdef CONFIG_RANDOM32_SELFTEST
static void __init prandom_state_selftest(void);
#else
static inline void prandom_state_selftest(void)
{
}
#endif

static DEFINE_PER_CPU(struct rnd_state, net_rand_state) __latent_entropy;

/**
 *	prandom_u32_state - seeded pseudo-random number generator.
 *	@state: pointer to state structure holding seeded state.
 *
 *	This is used for pseudo-randomness with no outside seeding.
 *	For more random results, use prandom_u32().
 */
u32 prandom_u32_state(struct rnd_state *state)
{
#define TAUSWORTHE(s, a, b, c, d) ((s & c) << d) ^ (((s << a) ^ s) >> b)
	state->s1 = TAUSWORTHE(state->s1,  6U, 13U, 4294967294U, 18U);
	state->s2 = TAUSWORTHE(state->s2,  2U, 27U, 4294967288U,  2U);
	state->s3 = TAUSWORTHE(state->s3, 13U, 21U, 4294967280U,  7U);
	state->s4 = TAUSWORTHE(state->s4,  3U, 12U, 4294967168U, 13U);

	return (state->s1 ^ state->s2 ^ state->s3 ^ state->s4);
}
EXPORT_SYMBOL(prandom_u32_state);

/**
 *	prandom_u32 - pseudo random number generator
 *
 *	A 32 bit pseudo-random number is generated using a fast
 *	algorithm suitable for simulation. This algorithm is NOT
 *	considered safe for cryptographic use.
 */
u32 prandom_u32(void)
{
	struct rnd_state *state = &get_cpu_var(net_rand_state);
	u32 res;

	res = prandom_u32_state(state);
	put_cpu_var(net_rand_state);

	return res;
}
EXPORT_SYMBOL(prandom_u32);

/**
 *	prandom_bytes_state - get the requested number of pseudo-random bytes
 *
 *	@state: pointer to state structure holding seeded state.
 *	@buf: where to copy the pseudo-random bytes to
 *	@bytes: the requested number of bytes
 *
 *	This is used for pseudo-randomness with no outside seeding.
 *	For more random results, use prandom_bytes().
 */
void prandom_bytes_state(struct rnd_state *state, void *buf, size_t bytes)
{
	u8 *ptr = buf;

	while (bytes >= sizeof(u32)) {
		put_unaligned(prandom_u32_state(state), (u32 *) ptr);
		ptr += sizeof(u32);
		bytes -= sizeof(u32);
	}

	if (bytes > 0) {
		u32 rem = prandom_u32_state(state);
		do {
			*ptr++ = (u8) rem;
			bytes--;
			rem >>= BITS_PER_BYTE;
		} while (bytes > 0);
	}
}
EXPORT_SYMBOL(prandom_bytes_state);

/**
 *	prandom_bytes - get the requested number of pseudo-random bytes
 *	@buf: where to copy the pseudo-random bytes to
 *	@bytes: the requested number of bytes
 */
void prandom_bytes(void *buf, size_t bytes)
{
	struct rnd_state *state = &get_cpu_var(net_rand_state);

	prandom_bytes_state(state, buf, bytes);
	put_cpu_var(net_rand_state);
}
EXPORT_SYMBOL(prandom_bytes);

static void prandom_warmup(struct rnd_state *state)
{
	/* Calling RNG ten times to satisfy recurrence condition */
	prandom_u32_state(state);
	prandom_u32_state(state);
	prandom_u32_state(state);
	prandom_u32_state(state);
	prandom_u32_state(state);
	prandom_u32_state(state);
	prandom_u32_state(state);
	prandom_u32_state(state);
	prandom_u32_state(state);
	prandom_u32_state(state);
}

static u32 __extract_hwseed(void)
{
	unsigned int val = 0;

	(void)(arch_get_random_seed_int(&val) ||
	       arch_get_random_int(&val));

	return val;
}

static void prandom_seed_early(struct rnd_state *state, u32 seed,
			       bool mix_with_hwseed)
{
#define LCG(x)	 ((x) * 69069U)	/* super-duper LCG */
#define HWSEED() (mix_with_hwseed ? __extract_hwseed() : 0)
	state->s1 = __seed(HWSEED() ^ LCG(seed),        2U);
	state->s2 = __seed(HWSEED() ^ LCG(state->s1),   8U);
	state->s3 = __seed(HWSEED() ^ LCG(state->s2),  16U);
	state->s4 = __seed(HWSEED() ^ LCG(state->s3), 128U);
}

/**
 *	prandom_seed - add entropy to pseudo random number generator
 *	@seed: seed value
 *
 *	Add some additional seeding to the prandom pool.
 */
void prandom_seed(u32 entropy)
{
	int i;
	/*
	 * No locking on the CPUs, but then somewhat random results are, well,
	 * expected.
	 */
	for_each_possible_cpu(i) {
		struct rnd_state *state = &per_cpu(net_rand_state, i);

		state->s1 = __seed(state->s1 ^ entropy, 2U);
		prandom_warmup(state);
	}
}
EXPORT_SYMBOL(prandom_seed);

/*
 *	Generate some initially weak seeding values to allow
 *	to start the prandom_u32() engine.
 */
static int __init prandom_init(void)
{
	int i;

	prandom_state_selftest();

	for_each_possible_cpu(i) {
		struct rnd_state *state = &per_cpu(net_rand_state, i);
		u32 weak_seed = (i + jiffies) ^ random_get_entropy();

		prandom_seed_early(state, weak_seed, true);
		prandom_warmup(state);
	}

	return 0;
}
core_initcall(prandom_init);

static void __prandom_timer(struct timer_list *unused);

static DEFINE_TIMER(seed_timer, __prandom_timer);

static void __prandom_timer(struct timer_list *unused)
{
	u32 entropy;
	unsigned long expires;

	get_random_bytes(&entropy, sizeof(entropy));
	prandom_seed(entropy);

	/* reseed every ~60 seconds, in [40 .. 80) interval with slack */
	expires = 40 + prandom_u32_max(40);
	seed_timer.expires = jiffies + msecs_to_jiffies(expires * MSEC_PER_SEC);

	add_timer(&seed_timer);
}

static void __init __prandom_start_seed_timer(void)
{
	seed_timer.expires = jiffies + msecs_to_jiffies(40 * MSEC_PER_SEC);
	add_timer(&seed_timer);
}

void prandom_seed_full_state(struct rnd_state __percpu *pcpu_state)
{
	int i;

	for_each_possible_cpu(i) {
		struct rnd_state *state = per_cpu_ptr(pcpu_state, i);
		u32 seeds[4];

		get_random_bytes(&seeds, sizeof(seeds));
		state->s1 = __seed(seeds[0],   2U);
		state->s2 = __seed(seeds[1],   8U);
		state->s3 = __seed(seeds[2],  16U);
		state->s4 = __seed(seeds[3], 128U);

		prandom_warmup(state);
	}
}
EXPORT_SYMBOL(prandom_seed_full_state);

/*
 *	Generate better values after random number generator
 *	is fully initialized.
 */
static void __prandom_reseed(bool late)
{
	unsigned long flags;
	static bool latch = false;
	static DEFINE_SPINLOCK(lock);

	/* Asking for random bytes might result in bytes getting
	 * moved into the nonblocking pool and thus marking it
	 * as initialized. In this case we would double back into
	 * this function and attempt to do a late reseed.
	 * Ignore the pointless attempt to reseed again if we're
	 * already waiting for bytes when the nonblocking pool
	 * got initialized.
	 */

	/* only allow initial seeding (late == false) once */
	if (!spin_trylock_irqsave(&lock, flags))
		return;

	if (latch && !late)
		goto out;

	latch = true;
	prandom_seed_full_state(&net_rand_state);
out:
	spin_unlock_irqrestore(&lock, flags);
}

void prandom_reseed_late(void)
{
	__prandom_reseed(true);
}

static int __init prandom_reseed(void)
{
	__prandom_reseed(false);
	__prandom_start_seed_timer();
	return 0;
}
late_initcall(prandom_reseed);

#ifdef CONFIG_RANDOM32_SELFTEST
static struct prandom_test1 {
	u32 seed;
	u32 result;
} test1[] = {
	{ 1U, 3484351685U },
	{ 2U, 2623130059U },
	{ 3U, 3125133893U },
	{ 4U,  984847254U },
};

static struct prandom_test2 {
	u32 seed;
	u32 iteration;
	u32 result;
} test2[] = {
	/* Test cases against taus113 from GSL library. */
	{  931557656U, 959U, 2975593782U },
	{ 1339693295U, 876U, 3887776532U },
	{ 1545556285U, 961U, 1615538833U },
	{  601730776U, 723U, 1776162651U },
	{ 1027516047U, 687U,  511983079U },
	{  416526298U, 700U,  916156552U },
	{ 1395522032U, 652U, 2222063676U },
	{  366221443U, 617U, 2992857763U },
	{ 1539836965U, 714U, 3783265725U },
	{  556206671U, 994U,  799626459U },
	{  684907218U, 799U,  367789491U },
	{ 2121230701U, 931U, 2115467001U },
	{ 1668516451U, 644U, 3620590685U },
	{  768046066U, 883U, 2034077390U },
	{ 1989159136U, 833U, 1195767305U },
	{  536585145U, 996U, 3577259204U },
	{ 1008129373U, 642U, 1478080776U },
	{ 1740775604U, 939U, 1264980372U },
	{ 1967883163U, 508U,   10734624U },
	{ 1923019697U, 730U, 3821419629U },
	{  442079932U, 560U, 3440032343U },
	{ 1961302714U, 845U,  841962572U },
	{ 2030205964U, 962U, 1325144227U },
	{ 1160407529U, 507U,  240940858U },
	{  635482502U, 779U, 4200489746U },
	{ 1252788931U, 699U,  867195434U },
	{ 1961817131U, 719U,  668237657U },
	{ 1071468216U, 983U,  917876630U },
	{ 1281848367U, 932U, 1003100039U },
	{  582537119U, 780U, 1127273778U },
	{ 1973672777U, 853U, 1071368872U },
	{ 1896756996U, 762U, 1127851055U },
	{  847917054U, 500U, 1717499075U },
	{ 1240520510U, 951U, 2849576657U },
	{ 1685071682U, 567U, 1961810396U },
	{ 1516232129U, 557U,    3173877U },
	{ 1208118903U, 612U, 1613145022U },
	{ 1817269927U, 693U, 4279122573U },
	{ 1510091701U, 717U,  638191229U },
	{  365916850U, 807U,  600424314U },
	{  399324359U, 702U, 1803598116U },
	{ 1318480274U, 779U, 2074237022U },
	{  697758115U, 840U, 1483639402U },
	{ 1696507773U, 840U,  577415447U },
	{ 2081979121U, 981U, 3041486449U },
	{  955646687U, 742U, 3846494357U },
	{ 1250683506U, 749U,  836419859U },
	{  595003102U, 534U,  366794109U },
	{   47485338U, 558U, 3521120834U },
	{  619433479U, 610U, 3991783875U },
	{  704096520U, 518U, 4139493852U },
	{ 1712224984U, 606U, 2393312003U },
	{ 1318233152U, 922U, 3880361134U },
	{  855572992U, 761U, 1472974787U },
	{   64721421U, 703U,  683860550U },
	{  678931758U, 840U,  380616043U },
	{  692711973U, 778U, 1382361947U },
	{  677703619U, 530U, 2826914161U },
	{   92393223U, 586U, 1522128471U },
	{ 1222592920U, 743U, 3466726667U },
	{  358288986U, 695U, 1091956998U },
	{ 1935056945U, 958U,  514864477U },
	{  735675993U, 990U, 1294239989U },
	{ 1560089402U, 897U, 2238551287U },
	{   70616361U, 829U,   22483098U },
	{  368234700U, 731U, 2913875084U },
	{   20221190U, 879U, 1564152970U },
	{  539444654U, 682U, 1835141259U },
	{ 1314987297U, 840U, 1801114136U },
	{ 2019295544U, 645U, 3286438930U },
	{  469023838U, 716U, 1637918202U },
	{ 1843754496U, 653U, 2562092152U },
	{  400672036U, 809U, 4264212785U },
	{  404722249U, 965U, 2704116999U },
	{  600702209U, 758U,  584979986U },
	{  519953954U, 667U, 2574436237U },
	{ 1658071126U, 694U, 2214569490U },
	{  420480037U, 749U, 3430010866U },
	{  690103647U, 969U, 3700758083U },
	{ 1029424799U, 937U, 3787746841U },
	{ 2012608669U, 506U, 3362628973U },
	{ 1535432887U, 998U,   42610943U },
	{ 1330635533U, 857U, 3040806504U },
	{ 1223800550U, 539U, 3954229517U },
	{ 1322411537U, 680U, 3223250324U },
	{ 1877847898U, 945U, 2915147143U },
	{ 1646356099U, 874U,  965988280U },
	{  805687536U, 744U, 4032277920U },
	{ 1948093210U, 633U, 1346597684U },
	{  392609744U, 783U, 1636083295U },
	{  690241304U, 770U, 1201031298U },
	{ 1360302965U, 696U, 1665394461U },
	{ 1220090946U, 780U, 1316922812U },
	{  447092251U, 500U, 3438743375U },
	{ 1613868791U, 592U,  828546883U },
	{  523430951U, 548U, 2552392304U },
	{  726692899U, 810U, 1656872867U },
	{ 1364340021U, 836U, 3710513486U },
	{ 1986257729U, 931U,  935013962U },
	{  407983964U, 921U,  728767059U },
};

static void __init prandom_state_selftest(void)
{
	int i, j, errors = 0, runs = 0;
	bool error = false;

	for (i = 0; i < ARRAY_SIZE(test1); i++) {
		struct rnd_state state;

		prandom_seed_early(&state, test1[i].seed, false);
		prandom_warmup(&state);

		if (test1[i].result != prandom_u32_state(&state))
			error = true;
	}

	if (error)
		pr_warn("prandom: seed boundary self test failed\n");
	else
		pr_info("prandom: seed boundary self test passed\n");

	for (i = 0; i < ARRAY_SIZE(test2); i++) {
		struct rnd_state state;

		prandom_seed_early(&state, test2[i].seed, false);
		prandom_warmup(&state);

		for (j = 0; j < test2[i].iteration - 1; j++)
			prandom_u32_state(&state);

		if (test2[i].result != prandom_u32_state(&state))
			errors++;

		runs++;
		cond_resched();
	}

	if (errors)
		pr_warn("prandom: %d/%d self tests failed\n", errors, runs);
	else
		pr_info("prandom: %d self tests passed\n", runs);
}
#endif
back to top