[SPDK] Errors in NVMF during parsing multi-element SGL

[SPDK] Errors in NVMF during parsing multi-element SGL

[Sasha Kotchubievsky]

[-- Attachment #1: Type: text/plain, Size: 2757 bytes --]

Hi,

I'm trying to submit a patch for fixing integer overflow issue in perf 
application introduced in v19.01.

"git push review" command runs unit tests under Valgrind and reports 
following errors:

==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371)
==11537==    by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl 
(rdma_ut.c:156)
==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x4E3A825: CU_run_all_tests (in 
/usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x407C0D: main (rdma_ut.c:352)
==11537==
==11537== Conditional jump or move depends on uninitialised value(s)
==11537==    at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374)
==11537==    by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl 
(rdma_ut.c:156)
==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x4E3A825: CU_run_all_tests (in 
/usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x407C0D: main (rdma_ut.c:352)
==11537==
==11537== Conditional jump or move depends on uninitialised value(s)
==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371)
==11537==    by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl 
(rdma_ut.c:172)
==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x4E3A825: CU_run_all_tests (in 
/usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x407C0D: main (rdma_ut.c:352)
==11537==
==11537== Conditional jump or move depends on uninitialised value(s)
==11537==    at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374)
==11537==    by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl 
(rdma_ut.c:172)
==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x4E3A825: CU_run_all_tests (in 
/usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x407C0D: main (rdma_ut.c:352)
==11537==
/mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl: 
*ERROR*: SGL length 0x40000 exceeds max io size 0x20000
/mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl: 
*ERROR*: In-capsule data length 0x1000 exceeds capsule length 0x0
/mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl: 
*ERROR*: In-capsule data length 0x2000 exceeds capsule length 0x1000

It looks like some fields "rdma_req" are uninitialized.

Best regards

Sasha

Re: [SPDK] Errors in NVMF during parsing multi-element SGL

[Sasha Kotchubievsky]

[-- Attachment #1: Type: text/plain, Size: 7317 bytes --]

Great. Thanks



I submitted my patch:

https://review.gerrithub.io/c/spdk/spdk/+/443996


On 2/8/2019 7:43 PM, Walker, Benjamin wrote:
> On Fri, 2019-02-08 at 12:41 +0200, Sasha Kotchubievsky wrote:
>> OK.
>>
>> Thanks
>>
>> Is https://review.gerrithub.io/c/spdk/spdk/+/443791/ going be merged to
>> 19.01.1 ?
> I just tagged it for 19.01.1
>
>> On 2/7/2019 8:21 PM, Howell, Seth wrote:
>>> Oh sorry I see what you are talking about now. That is an expected error.
>>> Some of the unit tests are negative ones intended to evoke failure messages
>>> like that. It won't cause the unittests to fail.
>>>
>>> -----Original Message-----
>>> From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Howell, Seth
>>> Sent: Thursday, February 7, 2019 11:08 AM
>>> To: Storage Performance Development Kit <spdk(a)lists.01.org>
>>> Subject: Re: [SPDK] Errors in NVMF during parsing multi-element SGL
>>>
>>> Hi Sasha,
>>>
>>> I am unfamiliar with that second error. Have you posted it before? If not,
>>> could you create a GitHub issue with the repro steps?
>>>
>>> Thanks,
>>>
>>> Seth
>>>
>>> -----Original Message-----
>>> From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha
>>> Kotchubievsky
>>> Sent: Thursday, February 7, 2019 10:54 AM
>>> To: spdk(a)lists.01.org
>>> Subject: Re: [SPDK] Errors in NVMF during parsing multi-element SGL
>>>
>>> Hi,
>>>
>>> The patch looks fine with me. I vote "+1" in Gerrit.
>>>
>>> Does it also fix:
>>>
>>> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request
>>> _parse_sgl:
>>> *ERROR*: SGL length 0x40000 exceeds max io size 0x20000
>>>
>>> ?
>>>
>>> Does the issue exist only on "master" or in 19.10 too ? Can it affect
>>> functionality or stability of 19.10?
>>>
>>> I follow procedure from "SPDK Development" document at
>>> https://spdk.io/development/#patch . Yes, I believe, those errors is reason
>>> for submitting failure. At least, only those errors I found in log file
>>> created during submit.
>>>
>>>
>>> Thanks
>>>
>>> Sasha
>>>
>>>
>>> On 2/7/2019 6:12 PM, Howell, Seth wrote:
>>>> Thanks for reporting this Sasha,
>>>>
>>>> https://review.gerrithub.io/#/c/spdk/spdk/+/443791/ fixed the issue for me
>>>> locally. I hadn't noticed this because  I don't have the same push hooks
>>>> installed in my machine. I am assuming that you are using the ones John M.
>>>> supplied on the mailing list?
>>>> But this is another indication that we need to improve the way we run the
>>>> unittests on Jenkins. This has been on our radar for a couple of weeks,
>>>> and Darek just got a patch through to enable setting all of the flags to
>>>> on in the unittest job we run per patch on Jenkins so regressions like
>>>> this shouldn't get through in the future.
>>>>
>>>> I forget, does the failure of the pre-hook stop you from submitting your
>>>> patch, or are you able to override it?
>>>>
>>>> Thanks,
>>>>
>>>> Seth
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha
>>>> Kotchubievsky
>>>> Sent: Thursday, February 7, 2019 6:15 AM
>>>> To: Storage Performance Development Kit <spdk(a)lists.01.org>
>>>> Subject: [SPDK] Errors in NVMF during parsing multi-element SGL
>>>>
>>>> Hi,
>>>>
>>>> I'm trying to submit a patch for fixing integer overflow issue in perf
>>>> application introduced in v19.01.
>>>>
>>>> "git push review" command runs unit tests under Valgrind and reports
>>>> following errors:
>>>>
>>>> ==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl
>>>> (rdma.c:1371) ==11537==    by 0x4062CF:
>>>> test_spdk_nvmf_rdma_request_parse_sgl
>>>> (rdma_ut.c:156)
>>>> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x4E3A825: CU_run_all_tests (in
>>>> /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537==
>>>> Conditional jump or move depends on uninitialised value(s) ==11537==
>>>> at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==
>>>> by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl
>>>> (rdma_ut.c:156)
>>>> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x4E3A825: CU_run_all_tests (in
>>>> /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537==
>>>> Conditional jump or move depends on uninitialised value(s) ==11537==
>>>> at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371) ==11537==
>>>> by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
>>>> (rdma_ut.c:172)
>>>> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x4E3A825: CU_run_all_tests (in
>>>> /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537==
>>>> Conditional jump or move depends on uninitialised value(s) ==11537==
>>>> at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==
>>>> by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
>>>> (rdma_ut.c:172)
>>>> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x4E3A825: CU_run_all_tests (in
>>>> /usr/lib64/libcunit.so.1.0.1)
>>>> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537==
>>>> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_reque
>>>> st_parse_sgl:
>>>> *ERROR*: SGL length 0x40000 exceeds max io size 0x20000
>>>> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_reque
>>>> st_parse_sgl:
>>>> *ERROR*: In-capsule data length 0x1000 exceeds capsule length 0x0
>>>> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_reque
>>>> st_parse_sgl:
>>>> *ERROR*: In-capsule data length 0x2000 exceeds capsule length 0x1000
>>>>
>>>> It looks like some fields "rdma_req" are uninitialized.
>>>>
>>>> Best regards
>>>>
>>>> Sasha
>>>>
>>>> _______________________________________________
>>>> SPDK mailing list
>>>> SPDK(a)lists.01.org
>>>> https://lists.01.org/mailman/listinfo/spdk
>>>> _______________________________________________
>>>> SPDK mailing list
>>>> SPDK(a)lists.01.org
>>>> https://lists.01.org/mailman/listinfo/spdk
>>> _______________________________________________
>>> SPDK mailing list
>>> SPDK(a)lists.01.org
>>> https://lists.01.org/mailman/listinfo/spdk
>>> _______________________________________________
>>> SPDK mailing list
>>> SPDK(a)lists.01.org
>>> https://lists.01.org/mailman/listinfo/spdk
>>> _______________________________________________
>>> SPDK mailing list
>>> SPDK(a)lists.01.org
>>> https://lists.01.org/mailman/listinfo/spdk
>> _______________________________________________
>> SPDK mailing list
>> SPDK(a)lists.01.org
>> https://lists.01.org/mailman/listinfo/spdk
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk

Re: [SPDK] Errors in NVMF during parsing multi-element SGL

[Walker, Benjamin]

[-- Attachment #1: Type: text/plain, Size: 7163 bytes --]

On Fri, 2019-02-08 at 12:41 +0200, Sasha Kotchubievsky wrote:
> OK.
> 
> Thanks
> 
> Is https://review.gerrithub.io/c/spdk/spdk/+/443791/ going be merged to 
> 19.01.1 ?

I just tagged it for 19.01.1

> 
> On 2/7/2019 8:21 PM, Howell, Seth wrote:
> > Oh sorry I see what you are talking about now. That is an expected error.
> > Some of the unit tests are negative ones intended to evoke failure messages
> > like that. It won't cause the unittests to fail.
> > 
> > -----Original Message-----
> > From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Howell, Seth
> > Sent: Thursday, February 7, 2019 11:08 AM
> > To: Storage Performance Development Kit <spdk(a)lists.01.org>
> > Subject: Re: [SPDK] Errors in NVMF during parsing multi-element SGL
> > 
> > Hi Sasha,
> > 
> > I am unfamiliar with that second error. Have you posted it before? If not,
> > could you create a GitHub issue with the repro steps?
> > 
> > Thanks,
> > 
> > Seth
> > 
> > -----Original Message-----
> > From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha
> > Kotchubievsky
> > Sent: Thursday, February 7, 2019 10:54 AM
> > To: spdk(a)lists.01.org
> > Subject: Re: [SPDK] Errors in NVMF during parsing multi-element SGL
> > 
> > Hi,
> > 
> > The patch looks fine with me. I vote "+1" in Gerrit.
> > 
> > Does it also fix:
> > 
> > /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request
> > _parse_sgl:
> > *ERROR*: SGL length 0x40000 exceeds max io size 0x20000
> > 
> > ?
> > 
> > Does the issue exist only on "master" or in 19.10 too ? Can it affect
> > functionality or stability of 19.10?
> > 
> > I follow procedure from "SPDK Development" document at 
> > https://spdk.io/development/#patch . Yes, I believe, those errors is reason
> > for submitting failure. At least, only those errors I found in log file
> > created during submit.
> > 
> > 
> > Thanks
> > 
> > Sasha
> > 
> > 
> > On 2/7/2019 6:12 PM, Howell, Seth wrote:
> > > Thanks for reporting this Sasha,
> > > 
> > > https://review.gerrithub.io/#/c/spdk/spdk/+/443791/ fixed the issue for me
> > > locally. I hadn't noticed this because  I don't have the same push hooks
> > > installed in my machine. I am assuming that you are using the ones John M.
> > > supplied on the mailing list?
> > > But this is another indication that we need to improve the way we run the
> > > unittests on Jenkins. This has been on our radar for a couple of weeks,
> > > and Darek just got a patch through to enable setting all of the flags to
> > > on in the unittest job we run per patch on Jenkins so regressions like
> > > this shouldn't get through in the future.
> > > 
> > > I forget, does the failure of the pre-hook stop you from submitting your
> > > patch, or are you able to override it?
> > > 
> > > Thanks,
> > > 
> > > Seth
> > > 
> > > 
> > > -----Original Message-----
> > > From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha
> > > Kotchubievsky
> > > Sent: Thursday, February 7, 2019 6:15 AM
> > > To: Storage Performance Development Kit <spdk(a)lists.01.org>
> > > Subject: [SPDK] Errors in NVMF during parsing multi-element SGL
> > > 
> > > Hi,
> > > 
> > > I'm trying to submit a patch for fixing integer overflow issue in perf
> > > application introduced in v19.01.
> > > 
> > > "git push review" command runs unit tests under Valgrind and reports
> > > following errors:
> > > 
> > > ==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl
> > > (rdma.c:1371) ==11537==    by 0x4062CF:
> > > test_spdk_nvmf_rdma_request_parse_sgl
> > > (rdma_ut.c:156)
> > > ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x4E3A825: CU_run_all_tests (in
> > > /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537==
> > > Conditional jump or move depends on uninitialised value(s) ==11537==
> > > at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==
> > > by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl
> > > (rdma_ut.c:156)
> > > ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x4E3A825: CU_run_all_tests (in
> > > /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537==
> > > Conditional jump or move depends on uninitialised value(s) ==11537==
> > > at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371) ==11537==
> > > by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
> > > (rdma_ut.c:172)
> > > ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x4E3A825: CU_run_all_tests (in
> > > /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537==
> > > Conditional jump or move depends on uninitialised value(s) ==11537==
> > > at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==
> > > by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
> > > (rdma_ut.c:172)
> > > ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x4E3A825: CU_run_all_tests (in
> > > /usr/lib64/libcunit.so.1.0.1)
> > > ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537==
> > > /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_reque
> > > st_parse_sgl:
> > > *ERROR*: SGL length 0x40000 exceeds max io size 0x20000
> > > /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_reque
> > > st_parse_sgl:
> > > *ERROR*: In-capsule data length 0x1000 exceeds capsule length 0x0
> > > /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_reque
> > > st_parse_sgl:
> > > *ERROR*: In-capsule data length 0x2000 exceeds capsule length 0x1000
> > > 
> > > It looks like some fields "rdma_req" are uninitialized.
> > > 
> > > Best regards
> > > 
> > > Sasha
> > > 
> > > _______________________________________________
> > > SPDK mailing list
> > > SPDK(a)lists.01.org
> > > https://lists.01.org/mailman/listinfo/spdk
> > > _______________________________________________
> > > SPDK mailing list
> > > SPDK(a)lists.01.org
> > > https://lists.01.org/mailman/listinfo/spdk
> > 
> > _______________________________________________
> > SPDK mailing list
> > SPDK(a)lists.01.org
> > https://lists.01.org/mailman/listinfo/spdk
> > _______________________________________________
> > SPDK mailing list
> > SPDK(a)lists.01.org
> > https://lists.01.org/mailman/listinfo/spdk
> > _______________________________________________
> > SPDK mailing list
> > SPDK(a)lists.01.org
> > https://lists.01.org/mailman/listinfo/spdk
> 
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk

Re: [SPDK] Errors in NVMF during parsing multi-element SGL

[Sasha Kotchubievsky]

[-- Attachment #1: Type: text/plain, Size: 6420 bytes --]

OK.

Thanks

Is https://review.gerrithub.io/c/spdk/spdk/+/443791/ going be merged to 
19.01.1 ?

On 2/7/2019 8:21 PM, Howell, Seth wrote:
> Oh sorry I see what you are talking about now. That is an expected error. Some of the unit tests are negative ones intended to evoke failure messages like that. It won't cause the unittests to fail.
>
> -----Original Message-----
> From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Howell, Seth
> Sent: Thursday, February 7, 2019 11:08 AM
> To: Storage Performance Development Kit <spdk(a)lists.01.org>
> Subject: Re: [SPDK] Errors in NVMF during parsing multi-element SGL
>
> Hi Sasha,
>
> I am unfamiliar with that second error. Have you posted it before? If not, could you create a GitHub issue with the repro steps?
>
> Thanks,
>
> Seth
>
> -----Original Message-----
> From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha Kotchubievsky
> Sent: Thursday, February 7, 2019 10:54 AM
> To: spdk(a)lists.01.org
> Subject: Re: [SPDK] Errors in NVMF during parsing multi-element SGL
>
> Hi,
>
> The patch looks fine with me. I vote "+1" in Gerrit.
>
> Does it also fix:
>
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: SGL length 0x40000 exceeds max io size 0x20000
>
> ?
>
> Does the issue exist only on "master" or in 19.10 too ? Can it affect functionality or stability of 19.10?
>
> I follow procedure from "SPDK Development" document at https://spdk.io/development/#patch . Yes, I believe, those errors is reason for submitting failure. At least, only those errors I found in log file created during submit.
>
>
> Thanks
>
> Sasha
>
>
> On 2/7/2019 6:12 PM, Howell, Seth wrote:
>> Thanks for reporting this Sasha,
>>
>> https://review.gerrithub.io/#/c/spdk/spdk/+/443791/ fixed the issue for me locally. I hadn't noticed this because  I don't have the same push hooks installed in my machine. I am assuming that you are using the ones John M. supplied on the mailing list?
>> But this is another indication that we need to improve the way we run the unittests on Jenkins. This has been on our radar for a couple of weeks, and Darek just got a patch through to enable setting all of the flags to on in the unittest job we run per patch on Jenkins so regressions like this shouldn't get through in the future.
>>
>> I forget, does the failure of the pre-hook stop you from submitting your patch, or are you able to override it?
>>
>> Thanks,
>>
>> Seth
>>
>>
>> -----Original Message-----
>> From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha
>> Kotchubievsky
>> Sent: Thursday, February 7, 2019 6:15 AM
>> To: Storage Performance Development Kit <spdk(a)lists.01.org>
>> Subject: [SPDK] Errors in NVMF during parsing multi-element SGL
>>
>> Hi,
>>
>> I'm trying to submit a patch for fixing integer overflow issue in perf application introduced in v19.01.
>>
>> "git push review" command runs unit tests under Valgrind and reports following errors:
>>
>> ==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl
>> (rdma.c:1371) ==11537==    by 0x4062CF:
>> test_spdk_nvmf_rdma_request_parse_sgl
>> (rdma_ut.c:156)
>> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x4E3A825: CU_run_all_tests (in
>> /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537==
>> Conditional jump or move depends on uninitialised value(s) ==11537==
>> at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==
>> by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl
>> (rdma_ut.c:156)
>> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x4E3A825: CU_run_all_tests (in
>> /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537==
>> Conditional jump or move depends on uninitialised value(s) ==11537==
>> at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371) ==11537==
>> by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
>> (rdma_ut.c:172)
>> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x4E3A825: CU_run_all_tests (in
>> /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537==
>> Conditional jump or move depends on uninitialised value(s) ==11537==
>> at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==
>> by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
>> (rdma_ut.c:172)
>> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x4E3A825: CU_run_all_tests (in
>> /usr/lib64/libcunit.so.1.0.1)
>> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537==
>> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl:
>> *ERROR*: SGL length 0x40000 exceeds max io size 0x20000
>> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl:
>> *ERROR*: In-capsule data length 0x1000 exceeds capsule length 0x0
>> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl:
>> *ERROR*: In-capsule data length 0x2000 exceeds capsule length 0x1000
>>
>> It looks like some fields "rdma_req" are uninitialized.
>>
>> Best regards
>>
>> Sasha
>>
>> _______________________________________________
>> SPDK mailing list
>> SPDK(a)lists.01.org
>> https://lists.01.org/mailman/listinfo/spdk
>> _______________________________________________
>> SPDK mailing list
>> SPDK(a)lists.01.org
>> https://lists.01.org/mailman/listinfo/spdk
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk

Re: [SPDK] Errors in NVMF during parsing multi-element SGL

[Howell, Seth]

[-- Attachment #1: Type: text/plain, Size: 5984 bytes --]

Oh sorry I see what you are talking about now. That is an expected error. Some of the unit tests are negative ones intended to evoke failure messages like that. It won't cause the unittests to fail.

-----Original Message-----
From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Howell, Seth
Sent: Thursday, February 7, 2019 11:08 AM
To: Storage Performance Development Kit <spdk(a)lists.01.org>
Subject: Re: [SPDK] Errors in NVMF during parsing multi-element SGL

Hi Sasha,

I am unfamiliar with that second error. Have you posted it before? If not, could you create a GitHub issue with the repro steps?

Thanks,

Seth

-----Original Message-----
From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha Kotchubievsky
Sent: Thursday, February 7, 2019 10:54 AM
To: spdk(a)lists.01.org
Subject: Re: [SPDK] Errors in NVMF during parsing multi-element SGL

Hi,

The patch looks fine with me. I vote "+1" in Gerrit.

Does it also fix:

/mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl:
*ERROR*: SGL length 0x40000 exceeds max io size 0x20000

?

Does the issue exist only on "master" or in 19.10 too ? Can it affect functionality or stability of 19.10?

I follow procedure from "SPDK Development" document at https://spdk.io/development/#patch . Yes, I believe, those errors is reason for submitting failure. At least, only those errors I found in log file created during submit.


Thanks

Sasha


On 2/7/2019 6:12 PM, Howell, Seth wrote:
> Thanks for reporting this Sasha,
>
> https://review.gerrithub.io/#/c/spdk/spdk/+/443791/ fixed the issue for me locally. I hadn't noticed this because  I don't have the same push hooks installed in my machine. I am assuming that you are using the ones John M. supplied on the mailing list?
> But this is another indication that we need to improve the way we run the unittests on Jenkins. This has been on our radar for a couple of weeks, and Darek just got a patch through to enable setting all of the flags to on in the unittest job we run per patch on Jenkins so regressions like this shouldn't get through in the future.
>
> I forget, does the failure of the pre-hook stop you from submitting your patch, or are you able to override it?
>
> Thanks,
>
> Seth
>
>
> -----Original Message-----
> From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha 
> Kotchubievsky
> Sent: Thursday, February 7, 2019 6:15 AM
> To: Storage Performance Development Kit <spdk(a)lists.01.org>
> Subject: [SPDK] Errors in NVMF during parsing multi-element SGL
>
> Hi,
>
> I'm trying to submit a patch for fixing integer overflow issue in perf application introduced in v19.01.
>
> "git push review" command runs unit tests under Valgrind and reports following errors:
>
> ==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl
> (rdma.c:1371) ==11537==    by 0x4062CF: 
> test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:156)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== 
> Conditional jump or move depends on uninitialised value(s) ==11537== 
> at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537== 
> by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:156)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== 
> Conditional jump or move depends on uninitialised value(s) ==11537== 
> at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371) ==11537== 
> by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:172)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== 
> Conditional jump or move depends on uninitialised value(s) ==11537== 
> at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537== 
> by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:172)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537==
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: SGL length 0x40000 exceeds max io size 0x20000
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: In-capsule data length 0x1000 exceeds capsule length 0x0
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: In-capsule data length 0x2000 exceeds capsule length 0x1000
>
> It looks like some fields "rdma_req" are uninitialized.
>
> Best regards
>
> Sasha
>
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk
_______________________________________________
SPDK mailing list
SPDK(a)lists.01.org
https://lists.01.org/mailman/listinfo/spdk
_______________________________________________
SPDK mailing list
SPDK(a)lists.01.org
https://lists.01.org/mailman/listinfo/spdk

Re: [SPDK] Errors in NVMF during parsing multi-element SGL

[Howell, Seth]

[-- Attachment #1: Type: text/plain, Size: 5406 bytes --]

Hi Sasha,

I am unfamiliar with that second error. Have you posted it before? If not, could you create a GitHub issue with the repro steps?

Thanks,

Seth

-----Original Message-----
From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha Kotchubievsky
Sent: Thursday, February 7, 2019 10:54 AM
To: spdk(a)lists.01.org
Subject: Re: [SPDK] Errors in NVMF during parsing multi-element SGL

Hi,

The patch looks fine with me. I vote "+1" in Gerrit.

Does it also fix:

/mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl:
*ERROR*: SGL length 0x40000 exceeds max io size 0x20000

?

Does the issue exist only on "master" or in 19.10 too ? Can it affect functionality or stability of 19.10?

I follow procedure from "SPDK Development" document at https://spdk.io/development/#patch . Yes, I believe, those errors is reason for submitting failure. At least, only those errors I found in log file created during submit.


Thanks

Sasha


On 2/7/2019 6:12 PM, Howell, Seth wrote:
> Thanks for reporting this Sasha,
>
> https://review.gerrithub.io/#/c/spdk/spdk/+/443791/ fixed the issue for me locally. I hadn't noticed this because  I don't have the same push hooks installed in my machine. I am assuming that you are using the ones John M. supplied on the mailing list?
> But this is another indication that we need to improve the way we run the unittests on Jenkins. This has been on our radar for a couple of weeks, and Darek just got a patch through to enable setting all of the flags to on in the unittest job we run per patch on Jenkins so regressions like this shouldn't get through in the future.
>
> I forget, does the failure of the pre-hook stop you from submitting your patch, or are you able to override it?
>
> Thanks,
>
> Seth
>
>
> -----Original Message-----
> From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha 
> Kotchubievsky
> Sent: Thursday, February 7, 2019 6:15 AM
> To: Storage Performance Development Kit <spdk(a)lists.01.org>
> Subject: [SPDK] Errors in NVMF during parsing multi-element SGL
>
> Hi,
>
> I'm trying to submit a patch for fixing integer overflow issue in perf application introduced in v19.01.
>
> "git push review" command runs unit tests under Valgrind and reports following errors:
>
> ==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl 
> (rdma.c:1371) ==11537==    by 0x4062CF: 
> test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:156)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== 
> Conditional jump or move depends on uninitialised value(s) ==11537==    
> at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==    
> by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:156)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== 
> Conditional jump or move depends on uninitialised value(s) ==11537==    
> at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371) ==11537==    
> by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:172)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== 
> Conditional jump or move depends on uninitialised value(s) ==11537==    
> at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==    
> by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:172)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) 
> ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537==
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: SGL length 0x40000 exceeds max io size 0x20000
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: In-capsule data length 0x1000 exceeds capsule length 0x0
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: In-capsule data length 0x2000 exceeds capsule length 0x1000
>
> It looks like some fields "rdma_req" are uninitialized.
>
> Best regards
>
> Sasha
>
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk
_______________________________________________
SPDK mailing list
SPDK(a)lists.01.org
https://lists.01.org/mailman/listinfo/spdk

Re: [SPDK] Errors in NVMF during parsing multi-element SGL

[Sasha Kotchubievsky]

[-- Attachment #1: Type: text/plain, Size: 4788 bytes --]

Hi,

The patch looks fine with me. I vote "+1" in Gerrit.

Does it also fix:

/mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl:
*ERROR*: SGL length 0x40000 exceeds max io size 0x20000

?

Does the issue exist only on "master" or in 19.10 too ? Can it affect 
functionality or stability of 19.10?

I follow procedure from "SPDK Development" document at 
https://spdk.io/development/#patch . Yes, I believe, those errors is 
reason for submitting failure. At least, only those errors I found in 
log file created during submit.


Thanks

Sasha


On 2/7/2019 6:12 PM, Howell, Seth wrote:
> Thanks for reporting this Sasha,
>
> https://review.gerrithub.io/#/c/spdk/spdk/+/443791/ fixed the issue for me locally. I hadn't noticed this because  I don't have the same push hooks installed in my machine. I am assuming that you are using the ones John M. supplied on the mailing list?
> But this is another indication that we need to improve the way we run the unittests on Jenkins. This has been on our radar for a couple of weeks, and Darek just got a patch through to enable setting all of the flags to on in the unittest job we run per patch on Jenkins so regressions like this shouldn't get through in the future.
>
> I forget, does the failure of the pre-hook stop you from submitting your patch, or are you able to override it?
>
> Thanks,
>
> Seth
>
>
> -----Original Message-----
> From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha Kotchubievsky
> Sent: Thursday, February 7, 2019 6:15 AM
> To: Storage Performance Development Kit <spdk(a)lists.01.org>
> Subject: [SPDK] Errors in NVMF during parsing multi-element SGL
>
> Hi,
>
> I'm trying to submit a patch for fixing integer overflow issue in perf application introduced in v19.01.
>
> "git push review" command runs unit tests under Valgrind and reports following errors:
>
> ==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371) ==11537==    by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:156)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== Conditional jump or move depends on uninitialised value(s) ==11537==    at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==    by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:156)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== Conditional jump or move depends on uninitialised value(s) ==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371) ==11537==    by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:172)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== Conditional jump or move depends on uninitialised value(s) ==11537==    at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==    by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
> (rdma_ut.c:172)
> ==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A825: CU_run_all_tests (in
> /usr/lib64/libcunit.so.1.0.1)
> ==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537==
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: SGL length 0x40000 exceeds max io size 0x20000
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: In-capsule data length 0x1000 exceeds capsule length 0x0
> /mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl:
> *ERROR*: In-capsule data length 0x2000 exceeds capsule length 0x1000
>
> It looks like some fields "rdma_req" are uninitialized.
>
> Best regards
>
> Sasha
>
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk
> _______________________________________________
> SPDK mailing list
> SPDK(a)lists.01.org
> https://lists.01.org/mailman/listinfo/spdk

Re: [SPDK] Errors in NVMF during parsing multi-element SGL

[Howell, Seth]

[-- Attachment #1: Type: text/plain, Size: 3900 bytes --]

Thanks for reporting this Sasha,

https://review.gerrithub.io/#/c/spdk/spdk/+/443791/ fixed the issue for me locally. I hadn't noticed this because  I don't have the same push hooks installed in my machine. I am assuming that you are using the ones John M. supplied on the mailing list?
But this is another indication that we need to improve the way we run the unittests on Jenkins. This has been on our radar for a couple of weeks, and Darek just got a patch through to enable setting all of the flags to on in the unittest job we run per patch on Jenkins so regressions like this shouldn't get through in the future.

I forget, does the failure of the pre-hook stop you from submitting your patch, or are you able to override it? 

Thanks,

Seth


-----Original Message-----
From: SPDK [mailto:spdk-bounces(a)lists.01.org] On Behalf Of Sasha Kotchubievsky
Sent: Thursday, February 7, 2019 6:15 AM
To: Storage Performance Development Kit <spdk(a)lists.01.org>
Subject: [SPDK] Errors in NVMF during parsing multi-element SGL

Hi,

I'm trying to submit a patch for fixing integer overflow issue in perf application introduced in v19.01.

"git push review" command runs unit tests under Valgrind and reports following errors:

==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371) ==11537==    by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl
(rdma_ut.c:156)
==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A825: CU_run_all_tests (in
/usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== Conditional jump or move depends on uninitialised value(s) ==11537==    at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==    by 0x4062CF: test_spdk_nvmf_rdma_request_parse_sgl
(rdma_ut.c:156)
==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A825: CU_run_all_tests (in
/usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== Conditional jump or move depends on uninitialised value(s) ==11537==    at 0x403E24: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1371) ==11537==    by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
(rdma_ut.c:172)
==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A825: CU_run_all_tests (in
/usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537== ==11537== Conditional jump or move depends on uninitialised value(s) ==11537==    at 0x403E56: spdk_nvmf_rdma_request_parse_sgl (rdma.c:1374) ==11537==    by 0x4065D5: test_spdk_nvmf_rdma_request_parse_sgl
(rdma_ut.c:172)
==11537==    by 0x4E3A211: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A4A7: ??? (in /usr/lib64/libcunit.so.1.0.1) ==11537==    by 0x4E3A825: CU_run_all_tests (in
/usr/lib64/libcunit.so.1.0.1)
==11537==    by 0x407C0D: main (rdma_ut.c:352) ==11537==
/mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1342:spdk_nvmf_rdma_request_parse_sgl: 
*ERROR*: SGL length 0x40000 exceeds max io size 0x20000
/mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl: 
*ERROR*: In-capsule data length 0x1000 exceeds capsule length 0x0
/mnt/data/sashakot/spdk_upstream/lib/nvmf/rdma.c:1404:spdk_nvmf_rdma_request_parse_sgl: 
*ERROR*: In-capsule data length 0x2000 exceeds capsule length 0x1000

It looks like some fields "rdma_req" are uninitialized.

Best regards

Sasha

_______________________________________________
SPDK mailing list
SPDK(a)lists.01.org
https://lists.01.org/mailman/listinfo/spdk