New
#101
I'm not sure how it works, mine is always showing around 800 (1600), I do know how it works if it was with the CPU but that is kinda basic stuff
I'm not sure how it works, mine is always showing around 800 (1600), I do know how it works if it was with the CPU but that is kinda basic stuff
I know this is an old post, but it is coming up at the top Google searches for e1r65x64.sys blue screens
I too am receiving a crash on the latest Intel NIC drivers. This is happening on a server with ECC memory running three instances of SQL. I have run a number of memory tests (not the extended ones that take hours) and cranked up SQL server's memory to consume it all, but the only time I have a BSOD is when the network traffic runs over 1Gbps for a long period of time.
The crashdump analysis shows that it always crashes in the NIC function TcpTcbReassemblyAdvanceHeader. Here is a snippet:
I also found that if I disable "Segment Coalescing" in the driver settings that the BSOD's seems to have subsided. At least for the last two days, I've been pushing 1-3Gbps continuously without issue.Code:DRIVER_IRQL_NOT_LESS_OR_EQUAL (d1) IMAGE_NAME: e1r65x64.sys STACK_TEXT: nt!KeBugCheckEx nt!KiBugCheckDispatch+0x69 nt!KiPageFault+0x426 tcpip!memcpy+0x2fa tcpip!TcpTcbReassemblyAdvanceHeader+0x110 tcpip!TcpTcbReassemblyRetrieveSegments+0x9d72b tcpip!TcpCanonicalizeUrgentTcbDatagram+0x69 tcpip!TcpTcbCarefulDatagram+0x382f tcpip!TcpTcbReceive+0x2e4 tcpip!TcpMatchReceive+0x1f0 tcpip!TcpPreValidatedReceive+0x3a2 tcpip!IppDeliverListToProtocol+0x93 tcpip!IppProcessDeliverList+0x62 tcpip!IppReceiveHeaderBatch+0x214 tcpip!IppFlcReceivePacketsCore+0x315 tcpip!FlpReceiveNonPreValidatedNetBufferListChain+0x271 tcpip!FlReceiveNetBufferListChainCalloutRoutine+0xc2 nt!KeExpandKernelStackAndCalloutInternal+0x85 tcpip!FlReceiveNetBufferListChain+0xb6 NDIS!ndisMIndicateNetBufferListsToOpen+0x11e NDIS!NdisMIndicateReceiveNetBufferLists+0x7e2
I also have "Large Send Offload V2" disabled, though I don't think this has any bearing.
Over the next couple of days, if it crashes again, I will probably stop back here and delete this reply with my apologizes. Or I'll be contacting Intel with the crashdump files I have.
Fingers crossed.
The Intel X540-T2 at that time.
This rabbit hole now goes deeper....
I purchased an Aquantia AQC-107 10GBase-T card to replace the X540. That new card has not caused any crashes since I replaced it.
However, internally there are two Intel 1Gbps NIC cards that are in use. Last night It just crashed in the same tcpip!TcbReassemblyAdvanceHeader+0x110 function in the e1r65x64.sys driver.
I am now convinced that this is a bug in the Intel drivers. Intel just released build 24.2 of their NIC drivers. I may try that, though I really doubt it's going to fix anything. Multi-threaded race conditions in drivers are not for the feint of heart.
I have a feeling that if I switch the internal NIC's to just one RSS queue, it will band-aid Intel's bug. However, that's hard to say, as this bug only seems to happen if I am using TWO Intel NIC's simultaneously on a multiple NUMA node system. Last night was the first time I was using both onboard at once. Most of the time it's using only one Intel NIC, and the AQC-107 NIC simultaneously, which does not cause an issue.
If the Intel driver mis-uses a static variable shared across all driver instances, that could explain why it only crashes when I push two Intel NIC's to their limits simultaneously on a multiple NUMA node system.
As I think about this, I realize that we push multiple Intel NIC's hard on a single NUMA node system, and there is never an issue. The crash happens on two different systems, both are multiple NUMA nodes, both using two Intel NIC's simultaneously.
(forgot to mention, I also swapped out the RAM just to be sure it wasn't the RAM. It's not the RAM.)
Last edited by Brain2000; 22 Sep 2019 at 14:50.
Please open a new thread.
See posting instructions:
BSOD - Posting Instructions