Revision 35bf49e054214a0c7d3b0a0f8606f753f3f6ae96 authored by Lu Baolu on 23 August 2022, 06:15:57 UTC, committed by Joerg Roedel on 07 September 2022, 13:14:57 UTC
With CONFIG_INTEL_IOMMU_DEBUGFS enabled, below lockdep splat are seen
when an I/O fault occurs on a machine with an Intel IOMMU in it.

 DMAR: DRHD: handling fault status reg 3
 DMAR: [DMA Write NO_PASID] Request device [00:1a.0] fault addr 0x0
       [fault reason 0x05] PTE Write access is not set
 DMAR: Dump dmar0 table entries for IOVA 0x0
 DMAR: root entry: 0x0000000127f42001
 DMAR: context entry: hi 0x0000000000001502, low 0x000000012d8ab001
 ================================
 WARNING: inconsistent lock state
 5.20.0-0.rc0.20220812git7ebfc85e2cd7.10.fc38.x86_64 #1 Not tainted
 --------------------------------
 inconsistent {HARDIRQ-ON-W} -> {IN-HARDIRQ-W} usage.
 rngd/1006 [HC1[1]:SC0[0]:HE0:SE1] takes:
 ff177021416f2d78 (&k->k_lock){?.+.}-{2:2}, at: klist_next+0x1b/0x160
 {HARDIRQ-ON-W} state was registered at:
   lock_acquire+0xce/0x2d0
   _raw_spin_lock+0x33/0x80
   klist_add_tail+0x46/0x80
   bus_add_device+0xee/0x150
   device_add+0x39d/0x9a0
   add_memory_block+0x108/0x1d0
   memory_dev_init+0xe1/0x117
   driver_init+0x43/0x4d
   kernel_init_freeable+0x1c2/0x2cc
   kernel_init+0x16/0x140
   ret_from_fork+0x1f/0x30
 irq event stamp: 7812
 hardirqs last  enabled at (7811): [<ffffffff85000e86>] asm_sysvec_apic_timer_interrupt+0x16/0x20
 hardirqs last disabled at (7812): [<ffffffff84f16894>] irqentry_enter+0x54/0x60
 softirqs last  enabled at (7794): [<ffffffff840ff669>] __irq_exit_rcu+0xf9/0x170
 softirqs last disabled at (7787): [<ffffffff840ff669>] __irq_exit_rcu+0xf9/0x170

The klist iterator functions using spin_*lock_irq*() but the klist
insertion functions using spin_*lock(), combined with the Intel DMAR
IOMMU driver iterating over klists from atomic (hardirq) context, where
pci_get_domain_bus_and_slot() calls into bus_find_device() which iterates
over klists.

As currently there's no plan to fix the klist to make it safe to use in
atomic context, this fixes the lockdep splat by avoid calling
pci_get_domain_bus_and_slot() in the hardirq context.

Fixes: 8ac0b64b9735 ("iommu/vt-d: Use pci_get_domain_bus_and_slot() in pgtable_walk()")
Reported-by: Lennert Buytenhek <buytenh@wantstofly.org>
Link: https://lore.kernel.org/linux-iommu/Yvo2dfpEh%2FWC+Wrr@wantstofly.org/
Link: https://lore.kernel.org/linux-iommu/YvyBdPwrTuHHbn5X@wantstofly.org/
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
Link: https://lore.kernel.org/r/20220819015949.4795-1-baolu.lu@linux.intel.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
1 parent a349ffc
Raw File
mst.c
// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * mst.c - NTFS multi sector transfer protection handling code. Part of the
 *	   Linux-NTFS project.
 *
 * Copyright (c) 2001-2004 Anton Altaparmakov
 */

#include "ntfs.h"

/**
 * post_read_mst_fixup - deprotect multi sector transfer protected data
 * @b:		pointer to the data to deprotect
 * @size:	size in bytes of @b
 *
 * Perform the necessary post read multi sector transfer fixup and detect the
 * presence of incomplete multi sector transfers. - In that case, overwrite the
 * magic of the ntfs record header being processed with "BAAD" (in memory only!)
 * and abort processing.
 *
 * Return 0 on success and -EINVAL on error ("BAAD" magic will be present).
 *
 * NOTE: We consider the absence / invalidity of an update sequence array to
 * mean that the structure is not protected at all and hence doesn't need to
 * be fixed up. Thus, we return success and not failure in this case. This is
 * in contrast to pre_write_mst_fixup(), see below.
 */
int post_read_mst_fixup(NTFS_RECORD *b, const u32 size)
{
	u16 usa_ofs, usa_count, usn;
	u16 *usa_pos, *data_pos;

	/* Setup the variables. */
	usa_ofs = le16_to_cpu(b->usa_ofs);
	/* Decrement usa_count to get number of fixups. */
	usa_count = le16_to_cpu(b->usa_count) - 1;
	/* Size and alignment checks. */
	if ( size & (NTFS_BLOCK_SIZE - 1)	||
	     usa_ofs & 1			||
	     usa_ofs + (usa_count * 2) > size	||
	     (size >> NTFS_BLOCK_SIZE_BITS) != usa_count)
		return 0;
	/* Position of usn in update sequence array. */
	usa_pos = (u16*)b + usa_ofs/sizeof(u16);
	/*
	 * The update sequence number which has to be equal to each of the
	 * u16 values before they are fixed up. Note no need to care for
	 * endianness since we are comparing and moving data for on disk
	 * structures which means the data is consistent. - If it is
	 * consistenty the wrong endianness it doesn't make any difference.
	 */
	usn = *usa_pos;
	/*
	 * Position in protected data of first u16 that needs fixing up.
	 */
	data_pos = (u16*)b + NTFS_BLOCK_SIZE/sizeof(u16) - 1;
	/*
	 * Check for incomplete multi sector transfer(s).
	 */
	while (usa_count--) {
		if (*data_pos != usn) {
			/*
			 * Incomplete multi sector transfer detected! )-:
			 * Set the magic to "BAAD" and return failure.
			 * Note that magic_BAAD is already converted to le32.
			 */
			b->magic = magic_BAAD;
			return -EINVAL;
		}
		data_pos += NTFS_BLOCK_SIZE/sizeof(u16);
	}
	/* Re-setup the variables. */
	usa_count = le16_to_cpu(b->usa_count) - 1;
	data_pos = (u16*)b + NTFS_BLOCK_SIZE/sizeof(u16) - 1;
	/* Fixup all sectors. */
	while (usa_count--) {
		/*
		 * Increment position in usa and restore original data from
		 * the usa into the data buffer.
		 */
		*data_pos = *(++usa_pos);
		/* Increment position in data as well. */
		data_pos += NTFS_BLOCK_SIZE/sizeof(u16);
	}
	return 0;
}

/**
 * pre_write_mst_fixup - apply multi sector transfer protection
 * @b:		pointer to the data to protect
 * @size:	size in bytes of @b
 *
 * Perform the necessary pre write multi sector transfer fixup on the data
 * pointer to by @b of @size.
 *
 * Return 0 if fixup applied (success) or -EINVAL if no fixup was performed
 * (assumed not needed). This is in contrast to post_read_mst_fixup() above.
 *
 * NOTE: We consider the absence / invalidity of an update sequence array to
 * mean that the structure is not subject to protection and hence doesn't need
 * to be fixed up. This means that you have to create a valid update sequence
 * array header in the ntfs record before calling this function, otherwise it
 * will fail (the header needs to contain the position of the update sequence
 * array together with the number of elements in the array). You also need to
 * initialise the update sequence number before calling this function
 * otherwise a random word will be used (whatever was in the record at that
 * position at that time).
 */
int pre_write_mst_fixup(NTFS_RECORD *b, const u32 size)
{
	le16 *usa_pos, *data_pos;
	u16 usa_ofs, usa_count, usn;
	le16 le_usn;

	/* Sanity check + only fixup if it makes sense. */
	if (!b || ntfs_is_baad_record(b->magic) ||
			ntfs_is_hole_record(b->magic))
		return -EINVAL;
	/* Setup the variables. */
	usa_ofs = le16_to_cpu(b->usa_ofs);
	/* Decrement usa_count to get number of fixups. */
	usa_count = le16_to_cpu(b->usa_count) - 1;
	/* Size and alignment checks. */
	if ( size & (NTFS_BLOCK_SIZE - 1)	||
	     usa_ofs & 1			||
	     usa_ofs + (usa_count * 2) > size	||
	     (size >> NTFS_BLOCK_SIZE_BITS) != usa_count)
		return -EINVAL;
	/* Position of usn in update sequence array. */
	usa_pos = (le16*)((u8*)b + usa_ofs);
	/*
	 * Cyclically increment the update sequence number
	 * (skipping 0 and -1, i.e. 0xffff).
	 */
	usn = le16_to_cpup(usa_pos) + 1;
	if (usn == 0xffff || !usn)
		usn = 1;
	le_usn = cpu_to_le16(usn);
	*usa_pos = le_usn;
	/* Position in data of first u16 that needs fixing up. */
	data_pos = (le16*)b + NTFS_BLOCK_SIZE/sizeof(le16) - 1;
	/* Fixup all sectors. */
	while (usa_count--) {
		/*
		 * Increment the position in the usa and save the
		 * original data from the data buffer into the usa.
		 */
		*(++usa_pos) = *data_pos;
		/* Apply fixup to data. */
		*data_pos = le_usn;
		/* Increment position in data as well. */
		data_pos += NTFS_BLOCK_SIZE/sizeof(le16);
	}
	return 0;
}

/**
 * post_write_mst_fixup - fast deprotect multi sector transfer protected data
 * @b:		pointer to the data to deprotect
 *
 * Perform the necessary post write multi sector transfer fixup, not checking
 * for any errors, because we assume we have just used pre_write_mst_fixup(),
 * thus the data will be fine or we would never have gotten here.
 */
void post_write_mst_fixup(NTFS_RECORD *b)
{
	le16 *usa_pos, *data_pos;

	u16 usa_ofs = le16_to_cpu(b->usa_ofs);
	u16 usa_count = le16_to_cpu(b->usa_count) - 1;

	/* Position of usn in update sequence array. */
	usa_pos = (le16*)b + usa_ofs/sizeof(le16);

	/* Position in protected data of first u16 that needs fixing up. */
	data_pos = (le16*)b + NTFS_BLOCK_SIZE/sizeof(le16) - 1;

	/* Fixup all sectors. */
	while (usa_count--) {
		/*
		 * Increment position in usa and restore original data from
		 * the usa into the data buffer.
		 */
		*data_pos = *(++usa_pos);

		/* Increment position in data as well. */
		data_pos += NTFS_BLOCK_SIZE/sizeof(le16);
	}
}
back to top