| Summary: | DAHLIN-00045: PCI Parity Error (Dell PowerEdge 1950 III) in spite of actual TE420p firmware (hardhdlc only) | ||
| Reporter: | Thomas Arimont (tomaso) | Labels: | |
| Date Opened: | 2008-09-23 03:42:44 | Date Closed: | 2008-10-01 15:31:13 |
| Priority: | Critical | Regression? | No |
| Status: | Closed/Complete | Components: | wct4xxp |
| Versions: | 2.0.0-rc2 | Frequency of Occurrence | |
| Related Issues: | |||
| Environment: | Attachments: | ||
| Description: | Our Dell PowerEdge 1950 III Server crashes in PCI Parity Error after one or two day of stress test (SIPP across 4x PRI loop) when d-channels in /etc/dahdi/system.conf are defined as 'hardhdlc'. TE420P firmware: ... TE4XXP version c01a016b, burst ON Octasic optimized! FALC version: 00000005, Board ID: 00 Reg 0: 0x30416400 Reg 1: 0x30416000 Reg 2: 0x00000000 Reg 3: 0x00000000 Reg 4: 0x00000101 Reg 5: 0x00000000 Reg 6: 0xc01a016b Reg 7: 0x00001f00 Reg 8: 0x010200ff Reg 9: 0x00fd0001 Reg 10: 0x0000004a ... Only after a few weeks testing we found out that hardhdlc configuration seems to cause these crashes (instead of dchan configuration) hardhdlc also causes the following asterisk/dahdi errors that we don' have in 'dchan'-mode: [Sep 23 12:49:46] ERROR[3607]: chan_dahdi.c:10107 dahdi_pri_error: !! Got reject for frame 120, retransmitting frame 122 now, updating n_r! [Sep 23 12:49:47] ERROR[3608]: chan_dahdi.c:10107 dahdi_pri_error: !! Got reject for frame 99, retransmitting frame 99 now, updating n_r! [Sep 23 12:49:47] ERROR[3608]: chan_dahdi.c:10107 dahdi_pri_error: !! Got reject for frame 99, retransmitting frame 100 now, updating n_r! [Sep 23 12:49:47] ERROR[3608]: chan_dahdi.c:10107 dahdi_pri_error: !! Got reject for frame 99, retransmitting frame 101 now, updating n_r! [Sep 23 12:49:47] ERROR[3608]: chan_dahdi.c:10107 dahdi_pri_error: !! Got reject for frame 99, retransmitting frame 102 now, updating n_r! [Sep 23 12:49:47] ERROR[3608]: chan_dahdi.c:10107 dahdi_pri_error: !! Got reject for frame 99, retransmitting frame 103 now, updating n_r! | ||
| Comments: | By: Leif Madsen (lmadsen) 2008-09-29 14:36:57 Is there a reason this is marked at private? By: Thomas Arimont (tomaso) 2008-09-30 00:44:11 Sorry, no, not really. Please change it. By the way, our examination (load tests) of this issue is still ongoing. The change of the hardhdlc-option to dchan-option could maybe only reduce the occurence of this crash because of less/different hardware register/memory mapped I/O access. We assume that the reason for the parity errors/system crashes is - since most commonly - an unsave PCI I/O handling (read/write transactions) of the dahdi driver. For example/info see http://lwn.net/Articles/125564/ By: Leif Madsen (lmadsen) 2008-10-01 10:41:35 Changed this back to public as it doesn't need to be private. Now everyone can help solve the issue! :) By: Russell Bryant (russell) 2008-10-01 15:31:10 Please contact Digium technical support | ||