I got Oops when using ext3-0.05d: enviroment and configure: official 2.2.18 kernel, with only kdb patch and ext3 patch contains in ext3-0.05d package, no reject or fuzz. compile ext3 direct into the kernel. Make a clean 200M ext2 partition and mount it, produce a 10M journal file, then remount it as ext3. Every operations is ok. Then umount it, repeat the same action( mke2fs, produce journal file, remount), and copy a large directory with many files to this partition, Oops occured: Eric kernel: Assertion failure in ext3_new_block() at balloc.c line 709: "!test_ and_set_bit(BH_Alloced, &bh->b_state)" Oops: 0002 CPU: 0 EIP: 0010:[<c0148d60>] EFLAGS: 00010246 eax: 0000006d ebx: 00000400 ecx: 0000004a edx: 0000002d esi: c7f8d000 edi: ffffffff ebp: c49d1dd4 esp: c49d1d70 ds: 0018 es: 0018 ss: 0018 Process cp (pid: 689, process nr: 27, stackpage=c49d1000) Stack: c01c877f 000002c5 c01c89a0 c3183a90 00000005 c3183a90 c3602f20 c3602f20 c0ccaca0 c3898020 c012bdd7 00000309 c37a9800 c36df800 00000000 c49d1dd0 00000112 c7f85000 c31da880 00000112 c3602f20 00000112 00000004 c31d5040 Call Trace: [<c01c877f>] [<c01c89a0>] [<c012bdd7>] [<c014b4e7>] [<c014ba6f>] [<c 014be06>] [<c0149a43>] [<c0121464>] [<c0121464>] [<c0121459>] [<c0121525>] [<c0121464>] [<c012a3 a4>] [<c0149720>] [<c010a24c>] Code: c6 05 00 00 00 00 00 83 c4 14 89 f6 8b 45 cc f0 0f ba 70 18
OK, looks as if some of the debugging flags are being carried over between mounts. I'll look into this. Is this repeatable?
Created attachment 7284 [details] Invalidate-buffers patch
Should be fixed by the attached patch, which I'll include in the next release.