Re: [PATCH v2 2/4] xfs_db: use iocursor type to guess btree geometry if bad magic

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 4/26/17 1:25 PM, Darrick J. Wong wrote:
> On Wed, Apr 26, 2017 at 12:48:48PM -0500, Eric Sandeen wrote:
>> On 4/10/17 7:20 PM, Darrick J. Wong wrote:
>>> From: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
>>>
>>> The function block_to_bt plays an integral role in determining the btree
>>> geometry of a block that we want to manipulate with the debugger.
>>> Normally we use the block magic to find the geometry profile, but if the
>>> magic is bad we'll never find it and return NULL.  The callers of this
>>> function do not check for NULL and crash.
>>>
>>> Therefore, if we can't find a geometry profile matching the magic
>>> number, use the iocursor type to guess the profile and scowl about that
>>> to stdout.  This makes it so that even with a corrupt magic we can try
>>> to print the fields instead of crashing the debugger.
>>>
>>> Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
>>
>> Makes sense generally, minor things below.
>>
>>> ---
>>> v2: be less macro-happy and only evaluate hascrc once
>>> v3: braces around the for loop body
>>> ---
>>>  db/btblock.c |   54 ++++++++++++++++++++++++++++++++++++++++++++++--------
>>>  1 file changed, 46 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/db/btblock.c b/db/btblock.c
>>> index 835a5f0..b819fa5 100644
>>> --- a/db/btblock.c
>>> +++ b/db/btblock.c
>>> @@ -25,6 +25,8 @@
>>>  #include "print.h"
>>>  #include "bit.h"
>>>  #include "init.h"
>>> +#include "io.h"
>>> +#include "output.h"
>>>  
>>>  /*
>>>   * Definition of the possible btree block layouts.
>>> @@ -122,13 +124,52 @@ static struct xfs_db_btree *
>>
>> urgh:
>>
>> /*
>>  * Find the right block defintion for a given ondisk block.
>>  *
>>  * We use the least significant bit of the magic number as index into
>>  * the array of block defintions.
>>
>> (nope, we don't, I can remove that on commit...)
> 
> Oops.  Should've nuked that comment from the start.

Ok, thanks for the replies.

I'll:

1) Fix up comments
2) /Not/ do backwards-jumping-gotos, fair point.
3) add the magic ASSERT
4) nuke the other ASSERTs

on commit? or do you want to resend ;)

> 
>>  */
>> static struct xfs_db_btree *
>>
>>>  block_to_bt(
>>>  	struct xfs_btree_block	*bb)
>>>  {
>>> -	struct xfs_db_btree *btp = &btrees[0];
>>> +	struct xfs_db_btree	*btp;
>>> +	uint32_t		magic;
>>> +	bool			crc;
>>>  
>>> -	do {
>>> -		if (be32_to_cpu((bb)->bb_magic) == btp->magic)
>>> +	magic = be32_to_cpu((bb)->bb_magic);
>>
>> maybe -
>>
>> again:
> 
> Eww, backwards jumping gotos :)
> 
> I guess that works so long as you're sure we can't accidentally jump
> backwards with an unrecognized magic and thus end up in an infinite
> loop.
> 
>>> +	for (btp = &btrees[0]; btp->magic != 0; btp++) {
>>> +		if (magic == btp->magic)
>>> +			return btp;
>>> +	}
>>
>> 	/* Magic is invalid/unknown.  Guess based on iocur type */
>>
>>> +	crc = xfs_sb_version_hascrc(&mp->m_sb);
>>
>> Bah, wish we could use something like:
>>
>> 	magic = xfs_magics[crc][btnum];
>>
>> but that's static to libxfs/xfs_btree.c and I guess btnum != typnm.
>> Because of course.
>>
>> (btnum = xfs_typnm_to_btnum[iocur_top->typ->typnm]; ? :) Ok I'll
>> drop it...)
> 
> :D
> 
>>> +	switch (iocur_top->typ->typnm) {
>>> +	case TYP_BMAPBTA:
>>> +	case TYP_BMAPBTD:
>>> +		magic = crc ? XFS_BMAP_CRC_MAGIC : XFS_BMAP_MAGIC;
>>> +		break;
>>> +	case TYP_BNOBT:
>>> +		magic = crc ? XFS_ABTB_CRC_MAGIC : XFS_ABTB_MAGIC;
>>> +		break;
>>> +	case TYP_CNTBT:
>>> +		magic = crc ? XFS_ABTC_CRC_MAGIC : XFS_ABTC_MAGIC;
>>> +		break;
>>> +	case TYP_INOBT:
>>> +		magic = crc ? XFS_IBT_CRC_MAGIC : XFS_IBT_MAGIC;
>>> +		break;
>>> +	case TYP_FINOBT:
>>> +		magic = crc ? XFS_FIBT_CRC_MAGIC : XFS_FIBT_MAGIC;
>>> +		break;
>>> +	case TYP_RMAPBT:
>>> +		magic = crc ? XFS_RMAP_CRC_MAGIC : 0;
>>> +		break;
>>> +	case TYP_REFCBT:
>>> +		magic = crc ? XFS_REFC_CRC_MAGIC : 0;
>>> +		break;
>>> +	default:
>>> +		ASSERT(0);
>>> +	}
>>
>> 	ASSERT(magic);
>>
>> we'd better have that by now, yes?
> 
> Yes.
> 
>> Or could we hit the RMAP/REFC types w/o crc set?
> 
> In theory, no, because xfs_db sets up a different type profile for the
> !crc case, and that type profile excludes TYP_RMAPBT/TYP_REFCBT.
> 
>>  I guess then the caller ASSERTs anyway?
> 
> <shrug>
> 
>>> +
>>> +	dbprintf(_("Bad btree magic 0x%x; coercing to %s.\n"),
>>> +		be32_to_cpu((bb)->bb_magic),
>>> +		iocur_top->typ->name);
>>
>>
>> 	goto again;
>>
>> to avoid the cut & paste of the loop below?  But maybe that's uglier.
> 
> Mostly a matter of taste, I think.  If the machinery was more complex
> I'd argue for a backwards goto to avoid having to maintain two identical
> loops, but this is pretty simple.
> 
>>> +
>>> +	for (btp = &btrees[0]; btp->magic != 0; btp++) {
>>> +		if (magic == btp->magic)
>>>  			return btp;
>>> -		btp++;
>>> -	} while (btp->magic != 0);
>>> +	}
>>>  
>>>  	return NULL;
>>>  }
>>> @@ -193,7 +234,6 @@ btblock_key_offset(
>>>  	int			offset;
>>>  
>>>  	ASSERT(startoff == 0);
>>> -	ASSERT(block->bb_level != 0);
>>
>> Help me out, why are these now removed?
> 
> /me doesn't remember, maybe they should just go away.
> 
>> -Eric
>>
>>>  
>>>  	offset = bt->block_len + (idx - 1) * bt->key_len;
>>>  	return bitize(offset);
>>> @@ -214,7 +254,6 @@ btblock_ptr_offset(
>>>  	int			maxrecs;
>>>  
>>>  	ASSERT(startoff == 0);
>>> -	ASSERT(block->bb_level != 0);
>>>  
>>>  	maxrecs = btblock_maxrecs(bt, mp->m_sb.sb_blocksize);
>>>  	offset = bt->block_len +
>>> @@ -238,7 +277,6 @@ btblock_rec_offset(
>>>  	int			offset;
>>>  
>>>  	ASSERT(startoff == 0);
>>> -	ASSERT(block->bb_level == 0);
>>>  
>>>  	offset = bt->block_len + (idx - 1) * bt->rec_len;
>>>  	return bitize(offset);
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
>>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [XFS Filesystem Development (older mail)]     [Linux Filesystem Development]     [Linux Audio Users]     [Yosemite Trails]     [Linux Kernel]     [Linux RAID]     [Linux SCSI]


  Powered by Linux