Skip to contentSkip to conversation

rcu: build on Windows

RFC In review Dharmik Thakkar 5 participants 9 messages Oct 19, 2020 – Oct 20, 2020

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

RFC 1/1 rcu: build on Windows

Build the lib for Windows.

Change-Id Ib83431a9ee56a7b38445f0474cb9a12b17cf18f9
---
 lib/meson.build | 1 +
 1 file changed, 1 insertion(+)
lib/meson.build +1 -0
index dd55b5cb53e4..1bb019720c6a 100644
@@ -41,6 +41,7 @@ if is_windows
 		'telemetry',
 		'eal',
 		'ring',
+		'rcu',
 		'mempool', 'mbuf', 'net', 'meter', 'ethdev', 'pci',
 		'cmdline',
 		'hash',
Signature · 1 line
2.17.1
#1 lore raw
Quoted 22 lines · Dharmik Thakkar

Build the lib for Windows.

Change-Id Ib83431a9ee56a7b38445f0474cb9a12b17cf18f9
---
 lib/meson.build | 1 +
 1 file changed, 1 insertion(+)
lib/meson.build +1 -1
index dd55b5cb53e4..1bb019720c6a 100644
@@ -41,6 +41,7 @@ if is_windows
                'telemetry',
                'eal',
                'ring',
+               'rcu',
                'mempool', 'mbuf', 'net', 'meter', 'ethdev', 'pci',
                'cmdline',
                'hash',
--
Quoted 2 lines

2.17.1

If we skip the new failures in Intel CI due to kni build being broken, I see that Windows build is fine but I have no log of it.
UNH lab skipped testing this patch, as it is a RFC (iiuc).
We need more tests, but on the principle, what do you think about enabling rcu?

Signature · 1 line
David Marchand
Tested-by
#2 lore raw
Quoted 22 lines · David Marchand
Quoted 21 lines · Dharmik Thakkar

Build the lib for Windows.

Change-Id Ib83431a9ee56a7b38445f0474cb9a12b17cf18f9
---
 lib/meson.build | 1 +
 1 file changed, 1 insertion(+)
lib/meson.build +1 -0
index dd55b5cb53e4..1bb019720c6a 100644
@@ -41,6 +41,7 @@ if is_windows
                'telemetry',
                'eal',
                'ring',
+               'rcu',
                'mempool', 'mbuf', 'net', 'meter', 'ethdev', 'pci',
                'cmdline',
                'hash',
Quoted 8 lines
Quoted 3 lines

--
2.17.1

If we skip the new failures in Intel CI due to kni build being broken, I see that Windows build is fine but I have no log of it.
UNH lab skipped testing this patch, as it is a RFC (iiuc).
We need more tests, but on the principle, what do you think about enabling rcu?

Builds OK with Linux/GCC (cross), Windows/MinGW-w64, Windows/clang.

#3 lore raw
Quoted 36 lines · Dmitry Kozlyuk
Quoted 22 lines · David Marchand
Quoted 21 lines · Dharmik Thakkar

Build the lib for Windows.

Change-Id Ib83431a9ee56a7b38445f0474cb9a12b17cf18f9

---
lib/meson.build | 1 +

1 file changed, 1 insertion(+)
lib/meson.build +1 -0
index dd55b5cb53e4..1bb019720c6a 100644
@@ -41,6 +41,7 @@ if is_windows
               'telemetry',
               'eal',
               'ring',
+               'rcu',
               'mempool', 'mbuf', 'net', 'meter', 'ethdev', 'pci',
               'cmdline',
               'hash',
Quoted 8 lines
Quoted 3 lines

--
2.17.1

If we skip the new failures in Intel CI due to kni build being broken, I see that Windows build is fine but I have no log of it.
UNH lab skipped testing this patch, as it is a RFC (iiuc).
We need more tests, but on the principle, what do you think about enabling rcu?

Quoted 3 lines

Builds OK with Linux/GCC (cross), Windows/MinGW-w64, Windows/clang.

Thank you for testing, Dmitry!

I tried cross-compilation on Linux (Ubuntu 18.04) with MinGW-w64 toolchain, but I am seeing some compilation error:

../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_probe’:
../lib/librte_eal/common/eal_common_dev.c:215:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?

    return -ENOMSG;
            ^~~~~~
            NPMSG
../lib/librte_eal/common/eal_common_dev.c:215:12: note: each undeclared identifier is reported only once for each function it appears in
../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_remove’:
../lib/librte_eal/common/eal_common_dev.c:361:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?
    return -ENOMSG;
            ^~~~~~
            NPMSG

[52/178] Compiling C object 'lib/lib@@rte_eal@sta/librte_eal_windows_eal_memory.c.obj'.
ninja: build stopped: subcommand failed.

Any help is appreciated.

#4 lore raw

<snip>

Quoted 63 lines
Quoted 2 lines

On Oct 20, 2020, at 3:34 AM, Dmitry Kozlyuk <[email protected]>

Quoted 3 lines · wrote:
Quoted 1 line · David Marchand

On Tue, Oct 20, 2020 at 1:34 AM Dharmik Thakkar

Quoted 21 lines · <[email protected]> wrote:
Quoted 21 lines
Quoted 21 lines

Build the lib for Windows.

Change-Id Ib83431a9ee56a7b38445f0474cb9a12b17cf18f9

---
lib/meson.build | 1 +

1 file changed, 1 insertion(+)
lib/meson.build index renamed from lib/meson.build +0 -0

dd55b5cb53e4..1bb019720c6a 100644

lib/meson.build +1 -0
@@ -41,6 +41,7 @@ if is_windows
               'telemetry',
               'eal',
               'ring',
+               'rcu',
               'mempool', 'mbuf', 'net', 'meter', 'ethdev', 'pci',
               'cmdline',
               'hash',
Quoted 8 lines
Quoted 8 lines
Quoted 3 lines

--
2.17.1

If we skip the new failures in Intel CI due to kni build being broken, I see that Windows build is fine but I have no log of it. UNH lab skipped testing this patch, as it is a RFC (iiuc). We need more tests, but on the principle, what do you think about

enabling rcu?

Quoted 4 lines

Builds OK with Linux/GCC (cross), Windows/MinGW-w64, Windows/clang.

Thank you for testing, Dmitry!

I tried cross-compilation on Linux (Ubuntu 18.04) with MinGW-w64 toolchain, but I am seeing some compilation error:

../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_probe’:
../lib/librte_eal/common/eal_common_dev.c:215:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?

    return -ENOMSG;
            ^~~~~~
            NPMSG

../lib/librte_eal/common/eal_common_dev.c:215:12: note: each undeclared identifier is reported only once for each function it appears in ../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_remove’:
../lib/librte_eal/common/eal_common_dev.c:361:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?

    return -ENOMSG;
            ^~~~~~
            NPMSG
Quoted 6 lines

[52/178] Compiling C object 'lib/lib@@rte_eal@sta/librte_eal_windows_eal_memory.c.obj'.
ninja: build stopped: subcommand failed.

Any help is appreciated.

Just to be clear, these errors are not stopping us from moving forward with this patch (as the CI for windows is passing and Dmitry has tested it in his environment).

Quoted 2 lines
#5 lore raw
In reply to Dharmik Thakkar
Quoted 14 lines · Dharmik Thakkar

I tried cross-compilation on Linux (Ubuntu 18.04) with MinGW-w64 toolchain, but I am seeing some compilation error:

../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_probe’:
../lib/librte_eal/common/eal_common_dev.c:215:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?

    return -ENOMSG;
            ^~~~~~
            NPMSG
../lib/librte_eal/common/eal_common_dev.c:215:12: note: each undeclared identifier is reported only once for each function it appears in
../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_remove’:
../lib/librte_eal/common/eal_common_dev.c:361:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?
    return -ENOMSG;
            ^~~~~~
            NPMSG
Quoted 5 lines

[52/178] Compiling C object 'lib/lib@@rte_eal@sta/librte_eal_windows_eal_memory.c.obj'.
ninja: build stopped: subcommand failed.

Any help is appreciated.

Ubuntu 18.04 has MinGW-w64 5.0.4, while 6.0.0 is required (note: MinGW-w64 version is not the same as GCC version). I've certainly hit this issue with older MinGW.

#6 lore raw
Quoted 23 lines · Dmitry Kozlyuk
Quoted 18 lines · Dharmik Thakkar

I tried cross-compilation on Linux (Ubuntu 18.04) with MinGW-w64 toolchain, but I am seeing some compilation error:

../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_probe’:
../lib/librte_eal/common/eal_common_dev.c:215:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?

   return -ENOMSG;
           ^~~~~~
           NPMSG
../lib/librte_eal/common/eal_common_dev.c:215:12: note: each undeclared identifier is reported only once for each function it appears in
../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_remove’:
../lib/librte_eal/common/eal_common_dev.c:361:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?
   return -ENOMSG;
           ^~~~~~
           NPMSG

[52/178] Compiling C object 'lib/lib@@rte_eal@sta/librte_eal_windows_eal_memory.c.obj'.
ninja: build stopped: subcommand failed.

Quoted 5 lines
Quoted 1 line

Any help is appreciated.

Ubuntu 18.04 has MinGW-w64 5.0.4, while 6.0.0 is required (note: MinGW-w64 version is not the same as GCC version). I've certainly hit this issue with older MinGW.

Got it, thank you! FWIW, documentation [1] doesn’t mention about the 6.0.0 requirement

[1] https://doc.dpdk.org/guides/windows_gsg/build_dpdk.html

#7 lore raw
Quoted 22 lines · Dharmik Thakkar
Quoted 22 lines · Dmitry Kozlyuk
Quoted 18 lines · Dharmik Thakkar

I tried cross-compilation on Linux (Ubuntu 18.04) with MinGW-w64 toolchain, but I am seeing some compilation error:

../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_probe’:
../lib/librte_eal/common/eal_common_dev.c:215:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?

    return -ENOMSG;
            ^~~~~~
            NPMSG
../lib/librte_eal/common/eal_common_dev.c:215:12: note: each undeclared identifier is reported only once for each function it appears in
../lib/librte_eal/common/eal_common_dev.c: In function ‘rte_dev_remove’:
../lib/librte_eal/common/eal_common_dev.c:361:12: error: ‘ENOMSG’ undeclared (first use in this function); did you mean ‘NPMSG’?
    return -ENOMSG;
            ^~~~~~
            NPMSG

[52/178] Compiling C object 'lib/lib@@rte_eal@sta/librte_eal_windows_eal_memory.c.obj'.
ninja: build stopped: subcommand failed.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

Any help is appreciated.

Ubuntu 18.04 has MinGW-w64 5.0.4, while 6.0.0 is required (note: MinGW-w64 version is not the same as GCC version). I've certainly hit this issue with older MinGW.

Got it, thank you! FWIW, documentation [1] doesn’t mention about the 6.0.0 requirement

[1] https://doc.dpdk.org/guides/windows_gsg/build_dpdk.html

While we're at this, we should also update this with the minimum required version of the Clang compiler.

ranjit m.

#8 lore raw
In reply to David Marchand
Quoted 26 lines · David Marchand
Quoted 21 lines · Dharmik Thakkar

Build the lib for Windows.

Change-Id Ib83431a9ee56a7b38445f0474cb9a12b17cf18f9

---
lib/meson.build | 1 +

1 file changed, 1 insertion(+)
lib/meson.build +1 -0
index dd55b5cb53e4..1bb019720c6a 100644
@@ -41,6 +41,7 @@ if is_windows
               'telemetry',
               'eal',
               'ring',
+               'rcu',
               'mempool', 'mbuf', 'net', 'meter', 'ethdev', 'pci',
               'cmdline',
               'hash',
Quoted 9 lines
Quoted 3 lines

--
2.17.1

If we skip the new failures in Intel CI due to kni build being broken, I see that Windows build is fine but I have no log of it.
UNH lab skipped testing this patch, as it is a RFC (iiuc).
We need more tests, but on the principle, what do you think about enabling rcu?

I think it should work OK.

Quoted 4 lines
Signature · 2 lines
-- 
David Marchand