SPDK CI downtime - July 29th - Aug 1st
by Latecki, Karol
Hi everyone! Just a short notice that SPDK Community CI will be offline during July 29th - August 1st weekend.
Why?
Scheduled electrical maintenance in the building.
For how long?
SPDK CI will be offline starting July 29th 11:00AM UTC time and will be back online on August 1st, 10:00AM UTC time.
How does that affect me?
SPDK CI will not pick up and test any patches sent to review.spdk.io during downtime.
Thanks,
Karol
3 weeks, 3 days
Re: The low water mark in NVMe/TCP target
by Walker, Benjamin
> From: Bin Yang <bin.yang(a)jaguarmicro.com>
>
> Hello, Everyone!
>
> I find spdk set low water mark in NVMe/TCP target to 8 bytes (sizeof(struct
> spdk_nvme_tcp_common_pdu_hdr)) In my test case, sometimes, there
> might be very small data packet (as small as 4 bytes) be sent to wire，After
> this, if there is no more data sent to the same socket, this small data packet
> won’t be received by NVMe/TCP controller qpair poll，
> because the size hasn’t reached the low watermark.
>
> So I want to know:
>
> 1) Why spdk need to set low watermark in NVMe/TCP target to 8 bytes，I
> don't think there's much advantage in doing that?
>
> 2) Can we set low watermark in NVMe/TCP target to 1 bytes just like iscsi
> target_does?
I think you're correct here - using 8 is a bug. It's absolutely possible that the last I/O sent was split into two weird data frame sizes that arrived significantly far apart in time, and the second one is something tiny like 4 bytes.
Did you want to submit a bug fix for this, or would you like me to?
Thanks,
Ben
>
> Thanks,
>
> Bin
> _______________________________________________
> SPDK mailing list -- spdk(a)lists.01.org
> To unsubscribe send an email to spdk-leave(a)lists.01.org
1 month
The low water mark in NVMe/TCP target
by Bin Yang
Hello, Everyone!
I find spdk set low water mark in NVMe/TCP target to 8 bytes (sizeof(struct spdk_nvme_tcp_common_pdu_hdr))
In my test case, sometimes, there might be very small data packet (as small as 4 bytes) be sent to wire，After
this, if there is no more data sent to the same socket, this small data packet won’t be received by NVMe/TCP controller qpair poll，
because the size hasn’t reached the low watermark.
So I want to know:
1) Why spdk need to set low watermark in NVMe/TCP target to 8 bytes，I don't think there's much advantage in doing that?
2) Can we set low watermark in NVMe/TCP target to 1 bytes just like iscsi target_does?
Thanks,
Bin
1 month, 1 week
SPDK CI downtime - July 12th
by Latecki, Karol
Hi everyone! Just a short notice that SPDK Community CI will be offline on July 12th.
Why?
Necessary updates and server maintenance.
For how long?
SPDK CI will be offline for a few hours an Tuesday July 12th, starting at 7:30AM GMT.
We estimate that downtime should not exceed 4 hours.
How does that affect me?
SPDK CI will not pick up and test any patches send to review.spdk.io during downtime.
Thanks,
Karol
1 month, 1 week
Reading data may cause dirty data（theory）
by 15735154041＠163.com
1. The process by spdk_nvme_ns_cmd_read read data
2. firmware write data to pcie address space
3. Process Exits unexpectedly（The firmware does not perceive that the process exits abnormally）
4. The pcie address space is allocated to another process
5. Firmware continues to write data to the pcie address space
6. another process read dirty data
SPDK reads data and converts host memory address (virtual address) to physical address (pcie address) in the driver layer.
After the firmware gets the physical address, it maps it to read and write
1 month, 2 weeks
Reading data may cause dirty data（theory）
by 15735154041＠163.com
1. The process by spdk_nvme_ns_cmd_read read data
2. firmware write data to pcie address space
3. Process Exits unexpectedly（The firmware does not perceive that the process exits abnormally）
4. The pcie address space is allocated to another process
5. Firmware continues to write data to the pcie address space
6. another process read dirty data
SPDK reads data and converts host memory address (virtual address) to physical address (pcie address) in the driver layer.
After the firmware gets the physical address, it maps it to read and write
1 month, 2 weeks
