diff mbox series

[v2] lib/cryptodev: fix assertion to remove GCC compilation warning

Message ID 20230522190453.453281-1-kamilx.godzwon@intel.com (mailing list archive)
State Changes Requested, archived
Delegated to: akhil goyal
Headers
Series [v2] lib/cryptodev: fix assertion to remove GCC compilation warning |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/loongarch-compilation success Compilation OK
ci/loongarch-unit-testing success Unit Testing PASS
ci/Intel-compilation success Compilation OK
ci/iol-mellanox-Performance success Performance Testing PASS
ci/intel-Testing success Testing PASS
ci/github-robot: build success github build: passed
ci/intel-Functional success Functional PASS
ci/iol-aarch64-unit-testing success Testing PASS
ci/iol-abi-testing success Testing PASS
ci/iol-aarch64-compile-testing success Testing PASS
ci/iol-unit-testing success Testing PASS
ci/iol-x86_64-compile-testing success Testing PASS
ci/iol-testing success Testing PASS
ci/iol-x86_64-unit-testing success Testing PASS
ci/iol-broadcom-Functional success Functional Testing PASS
ci/iol-intel-Functional success Functional Testing PASS
ci/iol-intel-Performance success Performance Testing PASS
ci/iol-broadcom-Performance success Performance Testing PASS

Commit Message

Kamil Godzwon May 22, 2023, 7:04 p.m. UTC 
  /home/vagrant/dpdk/build/include/rte_crypto_sym.h:1009:4: \
warning: Value stored to 'left' is never read [deadcode.DeadStores]
                          left = 0;
                          ^      ~
  1 warning generated.

Compilator sees that the variable 'left' is never read after
assignment a '0' value. To get rid of this warning message, use 'if'
condition to verify the 'left' value before RTE_ASSERT.

Signed-off-by: Kamil Godzwon <kamilx.godzwon@intel.com>
---
v2:
Changed commit message as the line was too long
Removed braces
---
 lib/cryptodev/rte_crypto_sym.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

Comments

Stephen Hemminger May 22, 2023, 9:25 p.m. UTC | #1 
On Mon, 22 May 2023 15:04:52 -0400
Kamil Godzwon <kamilx.godzwon@intel.com> wrote:

> /home/vagrant/dpdk/build/include/rte_crypto_sym.h:1009:4: \
> warning: Value stored to 'left' is never read [deadcode.DeadStores]
>                           left = 0;
>                           ^      ~
>   1 warning generated.
> 
> Compilator sees that the variable 'left' is never read after
> assignment a '0' value. To get rid of this warning message, use 'if'
> condition to verify the 'left' value before RTE_ASSERT.
> 
> Signed-off-by: Kamil Godzwon <kamilx.godzwon@intel.com>
> ---
> v2:
> Changed commit message as the line was too long
> Removed braces
> ---
>  lib/cryptodev/rte_crypto_sym.h | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/lib/cryptodev/rte_crypto_sym.h b/lib/cryptodev/rte_crypto_sym.h
> index b43174dbec..dcef1a5049 100644
> --- a/lib/cryptodev/rte_crypto_sym.h
> +++ b/lib/cryptodev/rte_crypto_sym.h
> @@ -1016,7 +1016,9 @@ rte_crypto_mbuf_to_vec(const struct rte_mbuf *mb, uint32_t ofs, uint32_t len,
>  		left -= seglen;
>  	}
>  
> -	RTE_ASSERT(left == 0);
> +	if (left != 0)
> +		RTE_ASSERT(false);
> +
>  	return i;
>  }
>  

This could happen if the passed in length to this routine was larger than
the amount of data in the mbuf. Should the function check and return an error?

Panic should only be reserved for seriously corrupted input (like invalid mbuf).

Also, this is a big enough function that it really should not be inlined.
Akhil Goyal May 23, 2023, 8:12 a.m. UTC | #2 
> On Mon, 22 May 2023 15:04:52 -0400
> Kamil Godzwon <kamilx.godzwon@intel.com> wrote:
> 
> > /home/vagrant/dpdk/build/include/rte_crypto_sym.h:1009:4: \
> > warning: Value stored to 'left' is never read [deadcode.DeadStores]
> >                           left = 0;
> >                           ^      ~
> >   1 warning generated.
> >
> > Compilator sees that the variable 'left' is never read after
> > assignment a '0' value. To get rid of this warning message, use 'if'
> > condition to verify the 'left' value before RTE_ASSERT.
> >
> > Signed-off-by: Kamil Godzwon <kamilx.godzwon@intel.com>
> > ---
> > v2:
> > Changed commit message as the line was too long
> > Removed braces
> > ---
> >  lib/cryptodev/rte_crypto_sym.h | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/lib/cryptodev/rte_crypto_sym.h b/lib/cryptodev/rte_crypto_sym.h
> > index b43174dbec..dcef1a5049 100644
> > --- a/lib/cryptodev/rte_crypto_sym.h
> > +++ b/lib/cryptodev/rte_crypto_sym.h
> > @@ -1016,7 +1016,9 @@ rte_crypto_mbuf_to_vec(const struct rte_mbuf
> *mb, uint32_t ofs, uint32_t len,
> >  		left -= seglen;
> >  	}
> >
> > -	RTE_ASSERT(left == 0);
> > +	if (left != 0)
> > +		RTE_ASSERT(false);
> > +
> >  	return i;
> >  }
> >
> 
> This could happen if the passed in length to this routine was larger than
> the amount of data in the mbuf. Should the function check and return an error?
> 
> Panic should only be reserved for seriously corrupted input (like invalid mbuf).
> 
> Also, this is a big enough function that it really should not be inlined.

This is a datapath API. RTE_ASSERT is normally not enabled in release build.
So, this assert is not doing any check for normal scenario.
We normally avoid these type of error checks in the datapath.
And while building in debug mode, we need these asserts to give a backtrace also
To debug the rootcause of the issue.

I would suggest fixing the assert itself instead of adding a check.
Current patch will affect performance.

Agreed, that the function is big for being an inline function,
but that is what all the datapath APIs are and
we keep them inline to improve the performance.
Stephen Hemminger May 23, 2023, 4 p.m. UTC | #3 
On Tue, 23 May 2023 08:12:28 +0000
Akhil Goyal <gakhil@marvell.com> wrote:

> > 
> > This could happen if the passed in length to this routine was larger than
> > the amount of data in the mbuf. Should the function check and return an error?
> > 
> > Panic should only be reserved for seriously corrupted input (like invalid mbuf).
> > 
> > Also, this is a big enough function that it really should not be inlined.  
> 
> This is a datapath API. RTE_ASSERT is normally not enabled in release build.
> So, this assert is not doing any check for normal scenario.
> We normally avoid these type of error checks in the datapath.
> And while building in debug mode, we need these asserts to give a backtrace also
> To debug the rootcause of the issue.
> 
> I would suggest fixing the assert itself instead of adding a check.
> Current patch will affect performance.
> 
> Agreed, that the function is big for being an inline function,
> but that is what all the datapath APIs are and
> we keep them inline to improve the performance.

Inline is not a magic go fast switch. Turns out that the compilers and cpu's
already do good job with functions.  Using LTO helps too.
Akhil Goyal May 24, 2023, 7:21 a.m. UTC | #4 
> On Tue, 23 May 2023 08:12:28 +0000
> Akhil Goyal <gakhil@marvell.com> wrote:
> 
> > >
> > > This could happen if the passed in length to this routine was larger than
> > > the amount of data in the mbuf. Should the function check and return an
> error?
> > >
> > > Panic should only be reserved for seriously corrupted input (like invalid
> mbuf).
> > >
> > > Also, this is a big enough function that it really should not be inlined.
> >
> > This is a datapath API. RTE_ASSERT is normally not enabled in release build.
> > So, this assert is not doing any check for normal scenario.
> > We normally avoid these type of error checks in the datapath.
> > And while building in debug mode, we need these asserts to give a backtrace
> also
> > To debug the rootcause of the issue.
> >
> > I would suggest fixing the assert itself instead of adding a check.
> > Current patch will affect performance.
> >
> > Agreed, that the function is big for being an inline function,
> > but that is what all the datapath APIs are and
> > we keep them inline to improve the performance.
> 
> Inline is not a magic go fast switch. Turns out that the compilers and cpu's
> already do good job with functions.  Using LTO helps too.

Inline perform better in cases of small functions.
For big functions like the above one, it may not be much beneficial,
but it won't cause any harm as well.
Yes, LTO can be explored. Thanks for the suggestion.
diff mbox series

Patch

diff --git a/lib/cryptodev/rte_crypto_sym.h b/lib/cryptodev/rte_crypto_sym.h
index b43174dbec..dcef1a5049 100644
--- a/lib/cryptodev/rte_crypto_sym.h
+++ b/lib/cryptodev/rte_crypto_sym.h
@@ -1016,7 +1016,9 @@  rte_crypto_mbuf_to_vec(const struct rte_mbuf *mb, uint32_t ofs, uint32_t len,
 		left -= seglen;
 	}
 
-	RTE_ASSERT(left == 0);
+	if (left != 0)
+		RTE_ASSERT(false);
+
 	return i;
 }