On Fri, Jan 05, 2018 at 01:31:08PM +0100, Jan Stary wrote:
> On Dec 25 22:07:12, o...@drijf.net wrote:
> > On Mon, Dec 25, 2017 at 11:48:22AM +0100, Jan Stary wrote:
> >
> > > This is current/amd64. The nightly dump (full mail and daily.local
> > > and df -hi at the bottom) reports a lot of errors such as
> > >
> > > read error from /dev/rsd2a: Invalid argument: [block -25232932862]:
> > > count=10240
> > >
> > > Obviously, there is no block -25232932862, but dump(8) must mean
> > > some certain block by that, if it knows the count.
> > > Is this simply an overflow in the (traverse.c)
> > >
> > > msg("read error from %s: %s: [block %lld]: count=%d\n",
> > > disk, strerror(errno), (long long)blkno, size);
> > >
> > > or the previous pread()? The disk seems to be working just fine,
> > > and a complete dd read of sd2c reports no errors:
> > >
> > > sd2:
> > > 3815602+1 records in
> > > 3815602+1 records out
> > > 250059350016 bytes transferred in 46028.063 secs (5432758 bytes/sec)
> >
> > I would start with unmounting the filesystem and doing a (forced) fsck
> > to see if your filesystem is corrupted.
>
> It's clean and fsck did not report anything.
>
> I should have mentioned that the failure happened during heavy activity
> on the disk (bitcoin full node synchronizing itself).
>
>
> On Jan 04 16:47:41, o...@drijf.net wrote:
> > On Thu, Jan 04, 2018 at 02:53:31PM +0100, Otto Moerbeek wrote:
> >
> > > On Thu, Jan 04, 2018 at 09:11:04AM +0100, Otto Moerbeek wrote:
> > >
> > > > On Wed, Jan 03, 2018 at 09:44:55PM -0600, Colton Lewis wrote:
> > > >
> > > > > When I try to run fsck on partition m of this disk:
> > > > >
> > > > > # /dev/rsd1c:
> > > > > type: SCSI
> > > > > disk: SCSI disk
> > > > > label: TOSHIBA MD04ACA4
> > > > > duid: 8ad0895bc1395d21
> > > > > flags:
> > > > > bytes/sector: 512
> > > > > sectors/track: 63
> > > > > tracks/cylinder: 255
> > > > > sectors/cylinder: 16065
> > > > > cylinders: 486401
> > > > > total sectors: 7814037168
> > > > > boundstart: 262208
> > > > > boundend: 7814037168
> > > > > drivedata: 0
> > > > >
> > > > > 16 partitions:
> > > > > # size offset fstype [fsize bsize cpg]
> > > > > a: 1136000 262208 4.2BSD 2048 16384 8875
> > > > > b: 1821490 1398208 swap
> > > > > c: 7814037168 0 unused
> > > > > d: 1571840 3219712 4.2BSD 2048 16384 12280
> > > > > e: 2318784 4791552 4.2BSD 2048 16384 12958
> > > > > f: 2672000 7110336 4.2BSD 2048 16384 12958
> > > > > g: 1545856 9782336 4.2BSD 2048 16384 12077
> > > > > h: 4944064 11328192 4.2BSD 2048 16384 12958
> > > > > i: 262144 64 MSDOS
> > > > > j: 2428672 16272256 4.2BSD 2048 16384 12958
> > > > > k: 6954496 18700928 4.2BSD 2048 16384 12958
> > > > > l: 7898912 25655424 4.2BSD 2048 16384 12958
> > > > > m: 7780482560 33554560 4.2BSD 8192 65536 1
> > > > >
> > > > > fsck reports that it cannot read negative block numbers:
> > > > >
> > > > > ** /dev/rsd1m
> > > > > BAD SUPER BLOCK: MAGIC NUMBER WRONG
> > > > >
> > > > > LOOK FOR ALTERNATE SUPERBLOCKS? yes
> > > > >
> > > > >
> > > > > CANNOT READ: BLK 749213312
> > > > > CONTINUE? yes
> > > > >
> > > > > THE FOLLOWING DISK SECTORS COULD NOT BE READ: 749213312, 749213313,
> > > > > 749213314, 749213315, 749213316, 749213317, 749213318, 749213319,
> > > > >
> > > > > CANNOT READ: BLK -2147483648
> > > > > CONTINUE? yes
> > > > >
> > > > > THE FOLLOWING DISK SECTORS COULD NOT BE READ: -2147483648,
> > > > > -2147483647, -2147483646, -2147483645, -2147483644, -2147483643,
> > > > > -2147483642, -2147483641, -2147483640, -2147483639, -2147483638,
> > > > > -2147483637, -2147483636, -2147483635, -2147483634, -2147483633,
> > > > >
> > > > > ...<repeat for the rest of the disk>
> > > > >
> > > > > How can I make sure fsck can handle a partition this size? There is
> > > > > nothing important on there at the moment.
> > > > >
> > > > > --
> > > > > Sincerely,
> > > > > Colton Lewis
> > > >
> > > > Did you actually newfs that partition? It looks like not since no
> > > > superblock or alternative is found.
> > > >
> > > > That said, it looks like there's an overflow somehere. I do not have
> > > > the hardware to investigate this though.
> > > >
> > > > On a side note: a partition that large will cause problem in other
> > > > areas. Even if it would work, the memory needed to do an fsck will be
> > > > huge.
> > > >
> > > > Also: provide dmeg! The platform involved can play a role in this.
> > > >
> > > > -Otto
> > >
> > > I tried to reproduce your problem using a vnd image using a sparse
> > > file.
> > >
> > > If I do not newfs the device, I get results very similar to what you
> > > are seeing.
> > >
> > > If I newfs the partition first, an fsck -f works as expected. So without
> > > further information, I assume you did not run newfs.
> > >
> > > I'll invetstigate the negative block numbers.
> > >
> > > -Otto
> >
> > THis diff should fixes the negative blocknumbers here,
>
> Given that my dump(8) problem above also reported negative block
> numbers, is there a similar glitch in dump? At some places, blkno
> is cast to different int_ types (but he disk code is way over my head).
Yest, that is very well possible.
-Otto
>
> Jan
>
> > Index: fsck.h
> > ===================================================================
> > RCS file: /cvs/src/sbin/fsck_ffs/fsck.h,v
> > retrieving revision 1.31
> > diff -u -p -r1.31 fsck.h
> > --- fsck.h 19 Jan 2015 18:20:47 -0000 1.31
> > +++ fsck.h 4 Jan 2018 15:46:37 -0000
> > @@ -229,7 +229,7 @@ extern long numdirs, listmax, inplast;
> > long secsize; /* actual disk sector size */
> > char nflag; /* assume a no response */
> > char yflag; /* assume a yes response */
> > -int bflag; /* location of alternate super block */
> > +daddr_t bflag; /* location of alternate super block */
> > int debug; /* output debugging info */
> > int cvtlevel; /* convert to newer file system format
> > */
> > char usedsoftdep; /* just fix soft dependency
> > inconsistencies */
> > Index: main.c
> > ===================================================================
> > RCS file: /cvs/src/sbin/fsck_ffs/main.c,v
> > retrieving revision 1.50
> > diff -u -p -r1.50 main.c
> > --- main.c 9 Sep 2016 15:37:15 -0000 1.50
> > +++ main.c 4 Jan 2018 15:46:37 -0000
> > @@ -48,7 +48,7 @@
> >
> > volatile sig_atomic_t returntosingle;
> >
> > -int argtoi(int, char *, char *, int);
> > +long long argtoi(int, char *, char *, int);
> > int checkfilesys(char *, char *, long, int);
> > int main(int, char *[]);
> >
> > @@ -78,7 +78,8 @@ main(int argc, char *argv[])
> > case 'b':
> > skipclean = 0;
> > bflag = argtoi('b', "number", optarg, 10);
> > - printf("Alternate super block location: %d\n", bflag);
> > + printf("Alternate super block location: %lld\n",
> > + (long long)bflag);
> > break;
> >
> > case 'c':
> > @@ -140,13 +141,13 @@ main(int argc, char *argv[])
> > exit(ret);
> > }
> >
> > -int
> > +long long
> > argtoi(int flag, char *req, char *str, int base)
> > {
> > char *cp;
> > - int ret;
> > + long long ret;
> >
> > - ret = (int)strtol(str, &cp, base);
> > + ret = strtoll(str, &cp, base);
> > if (cp == str || *cp)
> > errexit("-%c flag requires a %s\n", flag, req);
> > return (ret);
> > Index: setup.c
> > ===================================================================
> > RCS file: /cvs/src/sbin/fsck_ffs/setup.c,v
> > retrieving revision 1.63
> > diff -u -p -r1.63 setup.c
> > --- setup.c 9 Sep 2016 15:37:15 -0000 1.63
> > +++ setup.c 4 Jan 2018 15:46:37 -0000
> > @@ -202,7 +202,7 @@ setup(char *dev, int isfsdb)
> > }
> > found:
> > doskipclean = 0;
> > - pwarn("USING ALTERNATE SUPERBLOCK AT %d\n", bflag);
> > + pwarn("USING ALTERNATE SUPERBLOCK AT %lld\n", (long long)bflag);
> > }
> > if (debug)
> > printf("clean = %d\n", sblock.fs_clean);
> >
