I performed some R/W tests with a 2TB volume and it runs fine when exported from a POWER4 (PPC arch). vblade-21pre1.tar.gz was used. The current changelog is fine, thanks. .dan Em Monday 18 March 2013, Ed Cashin escreveu: > Thanks for the bugfix. > > Can you please build and run-time test the pre-release below, to verify > that it implements your fix? Also, please let me know if you don't want > your email address in the changelog. > > http://noserose.net/e/tmp/vblade-21pre1.tar.gz > > I'm attaching the updated patch, too, in case it gets through, after the > link to it below. > > http://noserose.net/e/tmp/vblade-20-21pre1.diff > > On Mar 4, 2013, at 3:53 PM, Daniel Mealha Cabrita wrote: > > hi there > > > > > > I've found and fixed a vblade (v20) bug which manifests itself in the > > following situation: > > > > - vblade host uses a big-endian processor (in our case, a POWER4). > > - exported volume has > 128GB (requires LBA48). > > > > The AoE client (in our case, Linux 2.6.32) ignores/gets_confused_with the > > LBA48 data and considers the LBA28 only, cropping the volume size to > > MAXLBA28SIZE-1. Some erratic behavior may happen: some partitions do not > > appear, unable to access root device etc. > > > > dmesg under Linux returns something like this: > > > > [5278516.938160] aoe: 00a1b0691845 e4.4 v4014 has 268435455 sectors > > [5278516.938216] etherd/e4.4: p1 p2 < > > > [5278516.938995] etherd/e4.4: p1 size 273874944 exceeds device capacity, > > limited to end of disk > > [5278516.939018] etherd/e4.4: p2 ignored, start 273879038 is behind the > > end of the disk > > > > > > .dan > > > > Following, the patch against vblade-20: > > > > --------------------------------------- > > diff -Nurp vblade-20/ata.c vblade-20_fixed_for_MSB_arch//ata.c > > --- vblade-20/ata.c 2009-08-14 15:33:09.000000000 -0300 > > +++ vblade-20_fixed_for_MSB_arch//ata.c 2013-03-04 17:39:44.000000000 > > -0300 @@ -23,16 +23,7 @@ enum { > > > > ERR = 1<<0, > > > > }; > > > > -static ushort ident[256] = { > > - [47] 0x8000, > > - [49] 0x0200, > > - [50] 0x4000, > > - [83] 0x5400, > > - [84] 0x4000, > > - [86] 0x1400, > > - [87] 0x4000, > > - [93] 0x400b, > > -}; > > +static ushort ident[256]; > > > > static void > > setfld(ushort *a, int idx, int len, char *str) // set field in ident > > @@ -79,6 +70,16 @@ setlba48(ushort *ident, vlong lba) > > > > *cp++ = lba >>= 8; > > *cp++ = lba >>= 8; > > > > } > > + > > +static void > > +setushort(ushort *a, int idx, ushort x) > > +{ > > + uchar *cp; > > + > > + cp = (uchar *)(a+idx); > > + *cp++ = x & 0xff; > > + *cp++ = x >> 8; > > +} > > > > void > > atainit(void) > > @@ -89,6 +90,15 @@ atainit(void) > > > > sprintf(buf, "V%d", VBLADE_VERSION); > > setfld(ident, 23, 8, buf); > > setfld(ident, 10, 20, serial); > > > > + > > + setushort(ident, 47, 0x8000); > > + setushort(ident, 49, 0x0200); > > + setushort(ident, 50, 0x4000); > > + setushort(ident, 83, 0x5400); > > + setushort(ident, 84, 0x4000); > > + setushort(ident, 86, 0x1400); > > + setushort(ident, 87, 0x4000); > > + setushort(ident, 93, 0x400b); > > } > > --------------------------------------- > > > > > > ------------------------------------------------------------------------- > > ----- Everyone hates slow websites. So do we. > > Make your web apps faster with AppDynamics > > Download AppDynamics Lite for free today: > > http://p.sf.net/sfu/appdyn_d2d_feb > > _______________________________________________ > > Aoetools-discuss mailing list > > Aoetools-discuss@xxxxxxxxxxxxxxxxxxxxx > > https://lists.sourceforge.net/lists/listinfo/aoetools-discuss > > -- > Ed Cashin > ecashin@xxxxxxxxxx ------------------------------------------------------------------------------ Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_d2d_mar _______________________________________________ Aoetools-discuss mailing list Aoetools-discuss@xxxxxxxxxxxxxxxxxxxxx https://lists.sourceforge.net/lists/listinfo/aoetools-discuss
