Bug 31379
| Summary: | Xircom Tulip CardBus Performance VERY Bad. | ||
|---|---|---|---|
| Product: | [Retired] Red Hat Linux | Reporter: | David Sainty <saint> |
| Component: | kernel | Assignee: | Arjan van de Ven <arjanv> |
| Status: | CLOSED WORKSFORME | QA Contact: | Brock Organ <borgan> |
| Severity: | medium | Docs Contact: | |
| Priority: | medium | ||
| Version: | 7.1 | ||
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | i386 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2003-06-06 00:00:59 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
htdig-3.1.5-6.src.rpm: 1.87 MB 4.61 MB/s htdig-3.2.0-0.b3.3.i386.rpm: 1.44 MB 2.04 MB/s htdig-web-3.2.0-0.b3.3.i386.rpm: 396.18 kB 4.76 MB/s kernel-2.2.16-22.i686.rpm: 6.69 MB 1.46 MB/s not quite 10MB/s, but not too bad here.... Can you try kernel 0.1.26 or later ? It has a New And Improved(tm) driver in it, still working on enhancing it even more. Arjan, Good news... I have got to the bottom of the performance problems we were seeing. With the xircom_tulip driver (both 2.4.2-0.1.22 and 2.4.2-0.1.29 kernels), this is the sort of performance we're seeing when connected to: - Cisco 4006; auto speed + duplex (Cisco picks 100FDX): 200kB/s + errors - " 100HDX: 8MB/s !!! - " 100FDX: 200kB/s + errors - " 10HDX: 65kB/s - " 10FDX: 80kB/s - to Cisco 4006 auto s+d VIA NetGear FS116 Fast Ether switch: 200kB/s - to Cisco 4006 VIA 3Com Office Connect Dual Speed Hub: 8MB/s It looks as though the only port setting on the Cisco that the card likes is 100, half duplex. 100 full duplex does NOT work (and 10 Mbit connections are bad also). Fix up FDX on the card and we have a winner! :-) I really, really hope you are talking about the xircom_cb driver and not xircom_tulip_cb :) I'll try to convince our local admin to play with my socket a bit to see if I can get this behavior as well. For me I cat ftp files at harddisk-speed :) Opps!! xircom_cb!! Sorry Arjan! ;-) Yeah, it will be interesting to see if you can repeat this behaviour. For me, as I said, its just a Cisco 4006 switch with 10/100 auto ports. Thanks. Our 3548 Cisco switches work fine on autodetect (they detect full duplex) Maybe your IOS needs updrading.... 12.0(5)XU works fine here. |
With the latest Xircom Tulip CardBus driver from Arjan / kernel, speed is _very_ slow, considering this Xircom Tulip is supposed to do 100Mbit FDX. Here are results from three tests conducted simultaneously, to receive rawhide RPMS off a server (e100 100FDX, RHL7 2.2.17 errata kernel) via a Cisco 4006 10/100 FDX capable switch using ncftp on the clients: Machine Card Av. Reported Transfer Speed ------------------------------------------------------------------------- 1. Latitude LS500 Xircom Tulip 10/100 CB card 185kB/sec 2. Latitude CPx650 3Com 3c574 10/100 PC (16b) card 800kB/sec 3. Dual PIII Xeon 3Com on-board 10/100 PCI 10MB/sec Considering this is supposed to be a CardBus 100Mbit FDX connection, the performance is 2% of what it should be. (Kernel used for 1. is 2.4.2-0.1.22, for 2. and 3. is 2.4.2-0.1.20) Also worth looking at may be the port statistics for 1. and 2 on the cisco: --------------------- From 1.... (the Xircom tulip CB card) RH-SURREY-SW0> (enable) show port 2/37 Port Name Status Vlan Level Duplex Speed Type ----- ------------------ ---------- ---------- ------ ------ ----- ------------ 2/37 connected 1 normal full 100 10/100BaseTX Port Security Violation Shutdown-Time Age-Time Max-Addr Trap IfIndex ----- -------- --------- ------------- -------- -------- -------- ------- 2/37 disabled shutdown 0 0 1 disabled 92 Port Num-Addr Secure-Src-Addr Age-Left Last-Src-Addr Shutdown/Time-Left ----- -------- ----------------- -------- ----------------- ------------------ 2/37 0 - - - - - Port Status Channel Admin Ch Mode Group Id ----- ---------- -------------------- ----- ----- 2/37 connected auto silent 24 0 Port Align-Err FCS-Err Xmit-Err Rcv-Err UnderSize ----- ---------- ---------- ---------- ---------- --------- 2/37 - 4573 0 9146 0 Port Single-Col Multi-Coll Late-Coll Excess-Col Carri-Sen Runts Giants ----- ---------- ---------- ---------- ---------- --------- --------- --------- 2/37 0 0 0 0 0 4573 0 Last-Time-Cleared -------------------------- Sat Mar 10 2001, 12:43:50 -------------- From 2... (3Com PC card) RH-SURREY-SW0> (enable) show port 2/33 Port Name Status Vlan Level Duplex Speed Type ----- ------------------ ---------- ---------- ------ ------ ----- ------------ 2/33 connected 1 normal full 100 10/100BaseTX Port Security Violation Shutdown-Time Age-Time Max-Addr Trap IfIndex ----- -------- --------- ------------- -------- -------- -------- ------- 2/33 disabled shutdown 0 0 1 disabled 88 Port Num-Addr Secure-Src-Addr Age-Left Last-Src-Addr Shutdown/Time-Left ----- -------- ----------------- -------- ----------------- ------------------ 2/33 0 - - - - - Port Status Channel Admin Ch Mode Group Id ----- ---------- -------------------- ----- ----- 2/33 connected auto silent 23 0 Port Align-Err FCS-Err Xmit-Err Rcv-Err UnderSize ----- ---------- ---------- ---------- ---------- --------- 2/33 - 17 0 933 900 Port Single-Col Multi-Coll Late-Coll Excess-Col Carri-Sen Runts Giants ----- ---------- ---------- ---------- ---------- --------- --------- --------- 2/33 0 0 0 0 0 16 0 Last-Time-Cleared -------------------------- Sat Mar 10 2001, 12:43:50 ------------------------- These two sets of port statistics were recorded at the same time, and the packet counts are over the same time period. Note the particularly high number of FCS errors, Rcv Errors and Runts on the port, even though there are no errors, dropped, overrun, etc when you do an ifconfig!! DS.. (ref #26060)