[ofa-general] 2.6.21 map problem

[Steve Wise]

Roland, do you know if anything changed in 2.6.21 regarding
remap_pfn_range()?  The chelsio bypass code is failing.  After mapping
the WQ memory that was allocated via dma_alloc_coherent(), the library
crashes when reading the mapped WQ memory.

The dma address being mapped is 0x7373a000.  It gets mapped to
0xaaaaaaad000. 

rping reads this address + 68 and we crash:

Mar  2 14:18:39 vic14 kernel: [78930.705415] iw_cxgb3: iwch_create_qp sq_num_entries 16, rq_num_entries 15 qpid 0x16 qhp ffff810078c3d488 dma_addr 0x7373a000 size 32
Mar  2 14:18:39 vic14 kernel: [78930.705520] iw_cxgb3: iwch_mmap pgoff 0x2 key 0x2000 len 4096
Mar  2 14:18:39 vic14 kernel: [78930.709966] iw_cxgb3: remove_mmap key 0x2000 addr 0xa800b000 len 4096
Mar  2 14:18:39 vic14 kernel: [78930.710047] iw_cxgb3: iwch_mmap pgoff 0x1 key 0x1000 len 4096
Mar  2 14:18:39 vic14 kernel: [78930.710122] iw_cxgb3: remove_mmap key 0x1000 addr 0x7373a000 len 4096
Mar  2 14:19:07 vic14 kernel: [78959.395682] rping: Corrupted page table at address 2aaaaaaad068
Mar  2 14:19:07 vic14 kernel: [78959.395759] PGD 1f3e6067 PUD 7201a067 PMD 22e27067 PTE 6b6b6b6b6b6b6027
Mar  2 14:19:07 vic14 kernel: [78959.396021] Bad pagetable: 000d [9] SMP
Mar  2 14:19:07 vic14 kernel: [78959.396182] CPU 2
Mar  2 14:19:07 vic14 kernel: [78959.396293] Modules linked in: iw_cxgb3 nfs lockd nfs_acl sunrpc rdma_krping rdma_ucm rdma_cm ib_cm iw_cm ib_addr ib_uverbs ib_umad ib_ipoib ib_sa ib_mthca ib_mad cxgb3 ib_core ipv6 af_packet button battery ac loop dm_mod e1000 parport_pc lp parport reiserfs edd fan thermal processor sg aic79xx scsi_transport_spi ata_piix libata piix sd_mod scsi_mod ide_disk ide_core
Mar  2 14:19:07 vic14 kernel: [78959.398617] Pid: 27938, comm: rping Not tainted 2.6.21-rc2 #1
Mar  2 14:19:07 vic14 kernel: [78959.398690] RIP: 0033:[<00002b5b5f5f0809>]  [<00002b5b5f5f0809>]
Mar  2 14:19:07 vic14 kernel: [78959.398813] RSP: 002b:00007fff4bc37980  EFLAGS: 00010302
Mar  2 14:19:07 vic14 kernel: [78959.398885] RAX: 00002aaaaaaad000 RBX: 00000000005093b0 RCX: 0000000000000000
Mar  2 14:19:07 vic14 kernel: [78959.398959] RDX: 0000000000000009 RSI: 00007fff4bc37aa0 RDI: 0000000000509458
Mar  2 14:19:07 vic14 kernel: [78959.399034] RBP: 00007fff4bc37980 R08: 0000000000000300 R09: 00002b5b5ef8cc00
Mar  2 14:19:07 vic14 kernel: [78959.399110] R10: 0000000000000000 R11: 00002b5b5f09c1f0 R12: 00007fff4bc37aa0
Mar  2 14:19:07 vic14 kernel: [78959.399184] R13: 00007fff4bc37d80 R14: 0000000000000000 R15: 0000000000000000
Mar  2 14:19:07 vic14 kernel: [78959.399260] FS:  00002b5b5f5eac50(0000) GS:ffff810001394af8(0000) knlGS:0000000000000000
Mar  2 14:19:07 vic14 kernel: [78959.399352] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
Mar  2 14:19:07 vic14 kernel: [78959.399425] CR2: 00002aaaaaaad068 CR3: 000000001cfc2000 CR4: 00000000000006e0
Mar  2 14:19:07 vic14 kernel: [78959.399500] Process rping (pid: 27938, threadinfo ffff810030aa8000, task ffff81007d882820)
Mar  2 14:19:07 vic14 kernel: [78959.399590]
Mar  2 14:19:07 vic14 kernel: [78959.399653] RIP  [<00002b5b5f5f0809>]
Mar  2 14:19:07 vic14 kernel: [78959.399769]  RSP <00007fff4bc37980>
Mar  2 14:23:09 vic14 syslog-ng[3158]: STATS: dropped 0


Any thoughts?

Thanks,

Steve.