XilSEM可插错错误类型及log示例

转载 于 2026-01-05 08:01:37 发布 · 2 阅读 · 0 ·
CC 4.0 BY-SA版权
原文链接：https://mp.weixin.qq.com/s?__biz=MzU1MDE0MjcxMQ==&mid=2247503775&idx=2&sn=23ed13d870f00932db786ed781cb1bae&chksm=fa3e1a37db3d80ce43278af1a35351db42cae6c282e0875e41710b315041a417ef8e3a46a99a&scene=126&sessionid=0 

关注、星标公众号，精彩内容每日送达
来源：网络素材

作者：Ivy Guo，AMD工程师；来源：AMD开发者社区

Versal的SEM解决方案XilSEM和之前的SEM IP有很大不同。SEM IP构建于FPGA的PL，基于底层硬核ICAP和FRAME_ECC进行扫描和纠正工作。而XilSEM是运行在PMC上的一个固件库，其插错，纠错等机制是预先编译的C代码。但两者检测到的bit错误类型是基本一致的。一般有CRAM存储单元的Correctable error, Uncorrectable error, CRC error等。另外，Versal系列器件中包含NPI寄存器，XilSEM的扫描范围同样覆盖到了NPI。

本文对几种错误类型做一下简单介绍，并基于VCK190开发板，熟悉一下发现错误后对应的XilSEM报告。

Correctable Error:

UltraSacle/+系列通过ECC策略和CRAM帧物理位置的设计，最多可以纠正一帧内16 bit错误。Versal不同于UltraScale+的结构, 一般单帧数据内按照只能纠正1-bit错误计算。所以在Versal里面测试可纠错误，我们选择单bit插错。可以在任意一个Row的一帧里面选择位置进行插错。一般有效的位置是Row:0~3，Qword:0~24， Bit:0~127。由于Qword12的bit:0~23，以及bit:48~71是syndrome数据，这些数据用于ECC校验，在这类位置插错不会影响用户逻辑，又可以验证插错效果，我们一般选择在此位置插错测试。

Source文件位于：

https://github.com/Xilinx/embeddedsw/blob/xlnx_rel_v2025.1/lib/sw_services/xilsem/examples/xsem_cram_example.c

用于插错的函数为：

XStatus Xsem_CfrApiInjctCorErr()

Example选择的插错位置为Frame 0，Qword 3，Bit 4，Row 0.

XSemCfrErrInjData ErrData = {0, 3, 4, 0};

使能XILSEM_CORR_ERRINJ_ENABLE宏。编译文件并下载到开发板上，得到的log如下：

CRAM scan is configured for immediate start

[XSem_CmdCfrReadFrameEcc] Success Reading Cfr Frame ECC Values of Row = 0x00  Frame = 0x0000000F  Over IPI

Received Segment 0 ECC Value = 0x00222A0A

Received Segment 1 ECC Value = 0x008A0200

[XSem_CfrApiStopScan] Success: Stop

Type: 0 Total Frames : 34111

Type: 1 Total Frames : 3520

Type: 2 Total Frames : 12800

Type: 3 Total Frames : 11

Type: 4 Total Frames : 5

Type: 5 Total Frames : 1

Type: 6 Total Frames : 0

Golden CRC for ROW_0 : 1BEB3B4E

[XSem_CfrApiStartScan] Success: Start

[XSem_CmdGetConfig] Success Reading CRAM and NPI configuration Over IPI

Received CRAM Configuration = 0x00000026

Received NPI Configuration = 0x00005012

[XSem_CfrApiStopScan] Success: Stop

starting Correctable Error injection

[XSem_CfrApiErrNjct] Success: Inject

[XSem_CfrApiStartScan] Success: Start

Received CRAM Correcatble error event

-----------------------------------------------------

-----------------Print Report------------------------

-----------------------------------------------------

[SUCCESS] Correctable error detected

[SUCCESS] Correction Done

Total Corrected Error count = 1

[SUCCESS] Error Count increased by 1 as expected

[SUCCESS] Received Correctable error event notification

Error is located in Row = 0, FAR = 0x000000, Qword = 3, Bit = 4

CRAM Scan Status:2014805

Uncorrectable Error:

当一帧内出错的bit超过1个时，一般就超出了XilSEM的纠错能力。此时XilSEM可以检测到错误的产生，但是无法纠正。为了模拟这一场景，我们可以在同一帧内插入2个错误：

   ErrData[0]: Frame Address : 0, Quadword: 0, Bit position: 2, Row: 0

   ErrData[1]: Frame Address : 0, Quadword: 0, Bit position: 4, Row: 0

错误地址为Frame 0，Qword 0，两个交错的bit：2和4，Row 0.

    XSemCfrErrInjData ErrData[2] = {{0, 0, 2, 0},

                                                          {0, 0, 4, 0},

                                                         };

此时得到如下log：

[XSem_CfrApiStartScan] Success: Start

[XSem_CmdGetConfig] Success Reading CRAM and NPI configuration Over IPI

Received CRAM Configuration = 0x00000026

Received NPI Configuration = 0x00005012

[XSem_CfrApiStopScan] Success: Stop

starting UnCorrectable Error injection

[XSem_CfrApiErrNjct] Success: Inject

[XSem_CfrApiErrNjct] Success: Inject

Received CRAM Uncorrectable error event

[XSem_CfrApiStartScan] Success: Start

-----------------------------------------------------

-----------------Print Report------------------------

-----------------------------------------------------

[SUCCESS] UnCorrectable error detected

[SUCCESS] No increase in Error Count

[SUCCESS] Received Uncorrectable error event notification

CRAM Scan Status:2010211

CRC Error：

如果想要模拟CRC错误，可以选择在同一个Qword内的交替位置，插入3个及以上的错误。此时会触发CRC校验错误。

位置：

     * ErrData[0]: Frame Address : 0, Quadword: 0, Bit position: 0, Row: 0

     * ErrData[1]: Frame Address : 0, Quadword: 0, Bit position: 2, Row: 0

     * ErrData[2]: Frame Address : 0, Quadword: 0, Bit position: 4, Row: 0

     * ErrData[3]: Frame Address : 0, Quadword: 0, Bit position: 6, Row: 0

CRC错误由于无法纠正，XilSEM在做了必要报告之后，会和Uncorrectable Err的处理一样，进入IDLE状态。

对应函数;

XSemCfrErrInjData ErrData[4] = {{0, 0, 0, 0},

                                                      {0, 0, 2, 0},

                                                      {0, 0, 4, 0},

                                                      {0, 0, 6, 0},

                                                    };

VCK190上执行的log如下：

CRAM scan is configured for immediate start

[XSem_CmdCfrReadFrameEcc] Success Reading Cfr Frame ECC Values of Row = 0x00  Frame = 0x0000000F  Over IPI

Received Segment 0 ECC Value = 0x00222A0A

Received Segment 1 ECC Value = 0x008A0200

[XSem_CfrApiStopScan] Success: Stop

Type: 0 Total Frames : 34111

Type: 1 Total Frames : 3520

Type: 2 Total Frames : 12800

Type: 3 Total Frames : 11

Type: 4 Total Frames : 5

Type: 5 Total Frames : 1

Type: 6 Total Frames : 0

Golden CRC for ROW_0 : 1BEB3B4E

[XSem_CfrApiStartScan] Success: Start

[XSem_CmdGetConfig] Success Reading CRAM and NPI configuration Over IPI

Received CRAM Configuration = 0x00000026

Received NPI Configuration = 0x00005012

[XSem_CfrApiStopScan] Success: Stop

starting UnCorrectable Error injection

[XSem_CfrApiErrNjct] Success: Inject

[XSem_CfrApiErrNjct] Success: Inject

[XSem_CfrApiErrNjct] Success: Inject

[XSem_CfrApiErrNjct] Success: Inject

[XSem_CfrApiStartScan] Success: Start

Received CRAM CRC error event

-----------------------------------------------------

-----------------Print Report------------------------

-----------------------------------------------------

[SUCCESS] CRC error detected

[SUCCESS] No increase in Error Count

[SUCCESS] Received CRC error event notification

CRAM Scan Status:2010411

可以看到，inject连续执行了四次，触发了CRC不可纠错误。

NPI Error：

在XilSEM库对NPI寄存器进行扫描时，采用了一种基于SHA（安全哈希算法，Secure Hash Algorithm）的检测策略。具体来说，XilSEM库会定期回读NPI寄存器的内容，并通过SHA算法计算其哈希值。通过将当前哈希值与预期或之前记录的哈希值进行比对，可以检测出寄存器内容是否发生了异常变化（即SEU错误）。因此要执行NPI扫描功能，必须要保证器件PMC的加密加速功能正常运行：

https://adaptivesupport.amd.com/s/article/000033536

Source文件位于：

https://github.com/Xilinx/embeddedsw/blob/xlnx_rel_v2025.1/lib/sw_services/xilsem/examples/xsem_npi_example.c

如Source文件中提到，Demo是往第一个描述符Descriptor的Golden SHA 中注入错误。在测试中，可以通过再次执行错误注入序列，纠正所注入的 SHA 错误。实际发生的 SHA 错误是不可纠正的。XilSEM在检测到错误后会做汇报，同时停止NPI扫描。CRAM扫描并不受此影响。

VCK190上执行的log如下：

-----------------------------------------

NPI Scan Fail Count for register 0: 0

NPI Scan Fail Count for register 1: 0

NPI Scan Fail Count for register 2: 0

NPI Scan Fail Count for register 3: 0

NPI Scan Fail Count for register 4: 0

NPI Scan Fail Count for register 5: 0

NPI Scan Fail Count for register 6: 0

NPI Scan Fail Count for register 7: 0

SHA mismatch Err details recorded as

ErrInfo[0]: 0

ErrInfo[1]: 0

NPI Scan Count = 5E85

HBCount = 266608

[main] Success: Get Golden SHA

[main] Success: Stop

[main] Success: Inject

[main] Success: Start

[main] Success: Get Golden SHA

-----------------------------------------------------

-----------------Print Report------------------------

-----------------------------------------------------

[SUCCESS] SHA comparison error detected

[SUCCESS] Scan counter not incrementing

[SUCCESS] Heartbeat counter not incrementing

[SUCCESS] Received CRC error event notification

----------------------------------------------------

Total Descriptor Count = 1

----------------------------------------------------

Descriptor information before injecting error:

Descriptor 1

  Type: Static

  Golden SHA: 0x16533699

----------------------------------------------------

Descriptor information after injecting error:

Descriptor 1

  Type: Static

  Golden SHA: 0x16533698

----------------------------------------------------

-------------- Test Report --------------

Failed Command Count : 0

NPI examples ran successfully

如果你的板子未开启加密加速模块，得到的log如下：

[ALERT] Received Cryptographic Accelerator Disabled event notification from XilSEM

[main] Success: Start

[main] ERROR: NPI Scan count not incrementing.

[main] Success: Get Golden SHA

[main] Success: Stop

[main] Success: Inject

[ALERT] Received Cryptographic Accelerator Disabled event notification from XilSEM

[main] Success: Start

[main] ERROR: Timeout occurred waiting for error.

-----------------------------------------------------

-----------------Print Report------------------------

-----------------------------------------------------

[SUCCESS] Scan counter not incrementing

[SUCCESS] Heartbeat counter not incrementing

[FAILURE] No CRC error event notification received

----------------------------------------------------

Total Descriptor Count = 1

----------------------------------------------------

Descriptor information before injecting error:

Descriptor 1

  Type: Static

  Golden SHA: 0x00000000

----------------------------------------------------

Descriptor information after injecting error:

Descriptor 1

  Type: Static

  Golden SHA: 0x00000000

----------------------------------------------------

-------------- Test Report --------------

Failed Command Count : 4

NPI examples Failed

以上是可以在实验室环境下进行测试的四种XilSEM错误类型。通过这些插错测试，可以熟悉一下XilSEM的报告和运行方式。其他如果是Fatal Error或者Internal Error，原因多样就无法模拟了。可以参考UG643：https://docs.amd.com/r/en-US/oslib_rm, Listening for SEU Detection做好系统中的应对措施。

（全文完）

声明：我们尊重原创，也注重分享；文字、图片版权归原作者所有。转载目的在于分享更多信息，不代表本号立场，如有侵犯您的权益请及时联系，我们将第一时间删除，谢谢！

图片

想要了解FPGA吗？这里有实例分享，ZYNQ设计，关注我们的公众号，探索

Hack电子
关注
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值