Re: Section conflict compile failures in net

Re: Section conflict compile failures in net

[James Bottomley]

On Fri, 2011-05-27 at 10:07 +0200, Michał Mirosław wrote:
> On Thu, May 26, 2011 at 04:39:53PM -0500, James Bottomley wrote:
> > I'm now getting a ton of errors like this in git head:
> > 
> >   CC [M]  drivers/net/3c59x.o
> >   CC [M]  drivers/net/hp100.o
> >   CC [M]  drivers/net/ne3210.o
> >   CC [M]  drivers/net/3c509.o
> >   CC [M]  drivers/net/depca.o
> > drivers/net/ne3210.c:83: error: irq_map causes a section type conflict
> > drivers/net/ne3210.c:85: error: shmem_map causes a section type conflict
> > drivers/net/ne3210.c:89: error: ifmap_val causes a section type conflict
> > drivers/net/ne3210.c:319: error: ne3210_ids causes a section type conflict
> > make[2]: *** [drivers/net/ne3210.o] Error 1
> > make[2]: *** Waiting for unfinished jobs....
> > drivers/net/hp100.c:198: error: hp100_eisa_tbl causes a section type conflict
> > drivers/net/hp100.c:211: error: hp100_pci_tbl causes a section type conflict
> > make[2]: *** [drivers/net/hp100.o] Error 1
> > drivers/net/depca.c:544: error: de1xx_irq causes a section type conflict
> > drivers/net/depca.c:545: error: de2xx_irq causes a section type conflict
> > drivers/net/depca.c:546: error: de422_irq causes a section type conflict
> [...]
> 
> Those three are only used in depca_hw_init() marked __devinit. What compiler
> [flags] do you use to build this?

It's a standard debian one.

jejb@ion> hppa64-linux-gnu-gcc -v
Using built-in specs.
Target: hppa64-linux-gnu
Configured with: ../src/configure --enable-languages=c --prefix=/usr
--libexecdir=/usr/lib --disable-shared --disable-nls --disable-threads
--disable-libffi --disable-libgomp --disable-libmudflap --disable-libssp
--with-as=/usr/bin/hppa64-linux-gnu-as
--with-ld=/usr/bin/hppa64-linux-gnu-ld
--includedir=/usr/hppa64-linux-gnu/include --host=hppa-linux-gnu
--build=hppa-linux-gnu --target=hppa64-linux-gnu
Thread model: single
gcc version 4.2.4 (Debian 4.2.4-6)

the problem is definitely the depca_irq[i] in the loop ... replace that
with a constant and the error goes away.

James

Re: Section conflict compile failures in net

[Michał Mirosław]

On Fri, May 27, 2011 at 06:04:37PM -0500, James Bottomley wrote:
> On Fri, 2011-05-27 at 10:07 +0200, Michał Mirosław wrote:
> > On Thu, May 26, 2011 at 04:39:53PM -0500, James Bottomley wrote:
> > > I'm now getting a ton of errors like this in git head:
> > > 
> > >   CC [M]  drivers/net/3c59x.o
> > >   CC [M]  drivers/net/hp100.o
> > >   CC [M]  drivers/net/ne3210.o
> > >   CC [M]  drivers/net/3c509.o
> > >   CC [M]  drivers/net/depca.o
> > > drivers/net/ne3210.c:83: error: irq_map causes a section type conflict
> > > drivers/net/ne3210.c:85: error: shmem_map causes a section type conflict
> > > drivers/net/ne3210.c:89: error: ifmap_val causes a section type conflict
> > > drivers/net/ne3210.c:319: error: ne3210_ids causes a section type conflict
> > > make[2]: *** [drivers/net/ne3210.o] Error 1
> > > make[2]: *** Waiting for unfinished jobs....
> > > drivers/net/hp100.c:198: error: hp100_eisa_tbl causes a section type conflict
> > > drivers/net/hp100.c:211: error: hp100_pci_tbl causes a section type conflict
> > > make[2]: *** [drivers/net/hp100.o] Error 1
> > > drivers/net/depca.c:544: error: de1xx_irq causes a section type conflict
> > > drivers/net/depca.c:545: error: de2xx_irq causes a section type conflict
> > > drivers/net/depca.c:546: error: de422_irq causes a section type conflict
> > [...]
> > 
> > Those three are only used in depca_hw_init() marked __devinit. What compiler
> > [flags] do you use to build this?
> 
> It's a standard debian one.
> 
> jejb@ion> hppa64-linux-gnu-gcc -v
> Using built-in specs.
> Target: hppa64-linux-gnu
> Configured with: ../src/configure --enable-languages=c --prefix=/usr
> --libexecdir=/usr/lib --disable-shared --disable-nls --disable-threads
> --disable-libffi --disable-libgomp --disable-libmudflap --disable-libssp
> --with-as=/usr/bin/hppa64-linux-gnu-as
> --with-ld=/usr/bin/hppa64-linux-gnu-ld
> --includedir=/usr/hppa64-linux-gnu/include --host=hppa-linux-gnu
> --build=hppa-linux-gnu --target=hppa64-linux-gnu
> Thread model: single
> gcc version 4.2.4 (Debian 4.2.4-6)
> 
> the problem is definitely the depca_irq[i] in the loop ... replace that
> with a constant and the error goes away.

Looks like arch-specific problem. Build test passes for x86 allyesconfig.

Best Regards,
Michał Mirosław

Re: Section conflict compile failures in net

[David Miller]

From: James Bottomley <James.Bottomley@HansenPartnership.com>
Date: Thu, 26 May 2011 16:39:53 -0500

BTW, linux-netdev doesn't exist, it's just plain netdev.

> Simply reverting 
> 
> commit e5cb966c0838e4da43a3b0751bdcac7fe719f7b4
> Author: Micha<C5><82> Miros<C5><82>aw <mirq-linux@rere.qmqm.pl>
> Date:   Mon Apr 18 13:31:20 2011 +0000
> 
>     net: fix section mismatches
>     
> Fixes the problem.
> 
> If I look at the first problem in hp100.c, the addition of
> __devinitconst to the device tables is fine, but there's no
> corresponding sectional tag on their use, so when compiled as a module,
> things like hp100_eisa_driver is now in the main data section but refers
> to something in the init data section ... I assume all the others are
> the same type of problem.

Yeah I think the device ID table __devinitconst bits have to be
reverted.

I'll apply something like the following:

net: Revert adding __devinitconst to driver ID tables.

If the table is hooked up to foo_driver->id_table we can't
mark it __devinitconst otherwise we end up with section
mismatches.

Reported-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/drivers/net/3c509.c b/drivers/net/3c509.c
index 5f25889..f6eaf6f 100644
--- a/drivers/net/3c509.c
+++ b/drivers/net/3c509.c
@@ -395,7 +395,7 @@ static struct isa_driver el3_isa_driver = {
 static int isa_registered;
 
 #ifdef CONFIG_PNP
-static const struct pnp_device_id el3_pnp_ids[] __devinitconst = {
+static const struct pnp_device_id el3_pnp_ids[] = {
 	{ .id = "TCM5090" }, /* 3Com Etherlink III (TP) */
 	{ .id = "TCM5091" }, /* 3Com Etherlink III */
 	{ .id = "TCM5094" }, /* 3Com Etherlink III (combo) */
@@ -478,7 +478,7 @@ static int pnp_registered;
 #endif /* CONFIG_PNP */
 
 #ifdef CONFIG_EISA
-static const struct eisa_device_id el3_eisa_ids[] __devinitconst = {
+static const struct eisa_device_id el3_eisa_ids[] = {
 		{ "TCM5090" },
 		{ "TCM5091" },
 		{ "TCM5092" },
@@ -508,7 +508,7 @@ static int eisa_registered;
 #ifdef CONFIG_MCA
 static int el3_mca_probe(struct device *dev);
 
-static const short el3_mca_adapter_ids[] __devinitconst = {
+static const short el3_mca_adapter_ids[] = {
 		0x627c,
 		0x627d,
 		0x62db,
diff --git a/drivers/net/3c59x.c b/drivers/net/3c59x.c
index 99f43d2..9ff7f37 100644
--- a/drivers/net/3c59x.c
+++ b/drivers/net/3c59x.c
@@ -901,7 +901,7 @@ static const struct dev_pm_ops vortex_pm_ops = {
 #endif /* !CONFIG_PM */
 
 #ifdef CONFIG_EISA
-static const struct eisa_device_id vortex_eisa_ids[] __devinitconst = {
+static const struct eisa_device_id vortex_eisa_ids[] = {
 	{ "TCM5920", CH_3C592 },
 	{ "TCM5970", CH_3C597 },
 	{ "" }
diff --git a/drivers/net/depca.c b/drivers/net/depca.c
index 1765405..645182b 100644
--- a/drivers/net/depca.c
+++ b/drivers/net/depca.c
@@ -342,7 +342,7 @@ static const char depca_string[] = "depca";
 static int depca_device_remove (struct device *device);
 
 #ifdef CONFIG_EISA
-static const struct eisa_device_id depca_eisa_ids[] __devinitconst = {
+static const struct eisa_device_id depca_eisa_ids[] = {
 	{ "DEC4220", de422 },
 	{ "" }
 };
diff --git a/drivers/net/hp100.c b/drivers/net/hp100.c
index c52a1df..ea8d345 100644
--- a/drivers/net/hp100.c
+++ b/drivers/net/hp100.c
@@ -195,7 +195,7 @@ static const char *const hp100_isa_tbl[] __devinitconst = {
 #endif
 
 #ifdef CONFIG_EISA
-static const struct eisa_device_id hp100_eisa_tbl[] __devinitconst = {
+static const struct eisa_device_id hp100_eisa_tbl[] = {
 	{ "HWPF180" }, /* HP J2577 rev A */
 	{ "HWP1920" }, /* HP 27248B */
 	{ "HWP1940" }, /* HP J2577 */
diff --git a/drivers/net/ibmlana.c b/drivers/net/ibmlana.c
index 136d754..00b1a2c 100644
--- a/drivers/net/ibmlana.c
+++ b/drivers/net/ibmlana.c
@@ -895,7 +895,7 @@ static int ibmlana_irq;
 static int ibmlana_io;
 static int startslot;		/* counts through slots when probing multiple devices */
 
-static const short ibmlana_adapter_ids[] __devinitconst = {
+static const short ibmlana_adapter_ids[] = {
 	IBM_LANA_ID,
 	0x0000
 };
diff --git a/drivers/net/irda/smsc-ircc2.c b/drivers/net/irda/smsc-ircc2.c
index 69b5707..76929e4 100644
--- a/drivers/net/irda/smsc-ircc2.c
+++ b/drivers/net/irda/smsc-ircc2.c
@@ -366,7 +366,7 @@ static inline void register_bank(int iobase, int bank)
 }
 
 /* PNP hotplug support */
-static const struct pnp_device_id smsc_ircc_pnp_table[] __devinitconst = {
+static const struct pnp_device_id smsc_ircc_pnp_table[] = {
 	{ .id = "SMCf010", .driver_data = 0 },
 	/* and presumably others */
 	{ }
diff --git a/drivers/net/ne3210.c b/drivers/net/ne3210.c
index e8984b0..3e7fb5b 100644
--- a/drivers/net/ne3210.c
+++ b/drivers/net/ne3210.c
@@ -316,7 +316,7 @@ static void ne3210_block_output(struct net_device *dev, int count,
 	memcpy_toio(shmem, buf, count);
 }
 
-static const struct eisa_device_id ne3210_ids[] __devinitconst = {
+static const struct eisa_device_id ne3210_ids[] = {
 	{ "EGL0101" },
 	{ "NVL1801" },
 	{ "" },
diff --git a/drivers/net/smc-mca.c b/drivers/net/smc-mca.c
index 0f29f26..76abd94 100644
--- a/drivers/net/smc-mca.c
+++ b/drivers/net/smc-mca.c
@@ -156,7 +156,7 @@ static const struct {
    { 14, 15 }
 };
 
-static const short smc_mca_adapter_ids[] __devinitconst = {
+static const short smc_mca_adapter_ids[] = {
 	0x61c8,
 	0x61c9,
 	0x6fc0,
diff --git a/drivers/net/tokenring/madgemc.c b/drivers/net/tokenring/madgemc.c
index 1313aa1..44b96f2 100644
--- a/drivers/net/tokenring/madgemc.c
+++ b/drivers/net/tokenring/madgemc.c
@@ -727,7 +727,7 @@ static int __devexit madgemc_remove(struct device *device)
 	return 0;
 }
 
-static const short madgemc_adapter_ids[] __devinitconst = {
+static const short madgemc_adapter_ids[] = {
 	0x002d,
 	0x0000
 };
diff --git a/drivers/net/tulip/de4x5.c b/drivers/net/tulip/de4x5.c
index 45144d5..40c0aa2 100644
--- a/drivers/net/tulip/de4x5.c
+++ b/drivers/net/tulip/de4x5.c
@@ -2097,7 +2097,7 @@ static int __devexit de4x5_eisa_remove (struct device *device)
 	return 0;
 }
 
-static const struct eisa_device_id de4x5_eisa_ids[] __devinitconst = {
+static const struct eisa_device_id de4x5_eisa_ids[] = {
         { "DEC4250", 0 },	/* 0 is the board name index... */
         { "" }
 };

Re: Section conflict compile failures in net

[Michał Mirosław]

On Tue, May 31, 2011 at 03:45:31PM -0700, David Miller wrote:
> From: James Bottomley <James.Bottomley@HansenPartnership.com>
> Date: Thu, 26 May 2011 16:39:53 -0500
> 
> BTW, linux-netdev doesn't exist, it's just plain netdev.
> 
> > Simply reverting 
> > 
> > commit e5cb966c0838e4da43a3b0751bdcac7fe719f7b4
> > Author: Micha<C5><82> Miros<C5><82>aw <mirq-linux@rere.qmqm.pl>
> > Date:   Mon Apr 18 13:31:20 2011 +0000
> > 
> >     net: fix section mismatches
> >     
> > Fixes the problem.
> > 
> > If I look at the first problem in hp100.c, the addition of
> > __devinitconst to the device tables is fine, but there's no
> > corresponding sectional tag on their use, so when compiled as a module,
> > things like hp100_eisa_driver is now in the main data section but refers
> > to something in the init data section ... I assume all the others are
> > the same type of problem.
> Yeah I think the device ID table __devinitconst bits have to be
> reverted.

__devinitconst ID tables do work for PCI drivers. What's different for EISA,
and why is it working in x86 allyesconfig build but not in hppa?

Best Regards,
Michał Mirosław

Re: Section conflict compile failures in net

[James Bottomley]

On Tue, 2011-05-31 at 15:45 -0700, David Miller wrote:
> From: James Bottomley <James.Bottomley@HansenPartnership.com>
> Date: Thu, 26 May 2011 16:39:53 -0500
> 
> BTW, linux-netdev doesn't exist, it's just plain netdev.

Um, yes ... I thought it was netdev, then I checked MARC and it had
linux-netdev.  I'll remember for next time (well, at least for the next
six months).

> > Simply reverting 
> > 
> > commit e5cb966c0838e4da43a3b0751bdcac7fe719f7b4
> > Author: Micha<C5><82> Miros<C5><82>aw <mirq-linux@rere.qmqm.pl>
> > Date:   Mon Apr 18 13:31:20 2011 +0000
> > 
> >     net: fix section mismatches
> >     
> > Fixes the problem.
> > 
> > If I look at the first problem in hp100.c, the addition of
> > __devinitconst to the device tables is fine, but there's no
> > corresponding sectional tag on their use, so when compiled as a module,
> > things like hp100_eisa_driver is now in the main data section but refers
> > to something in the init data section ... I assume all the others are
> > the same type of problem.
> 
> Yeah I think the device ID table __devinitconst bits have to be
> reverted.
> 
> I'll apply something like the following:
> 
> net: Revert adding __devinitconst to driver ID tables.
> 
> If the table is hooked up to foo_driver->id_table we can't
> mark it __devinitconst otherwise we end up with section
> mismatches.
> 
> Reported-by: James Bottomley <James.Bottomley@HansenPartnership.com>
> Signed-off-by: David S. Miller <davem@davemloft.net>

That fixes some but not all; the pci tables still cause a section
conflict:

drivers/net/hp100.c:211: error: hp100_pci_tbl causes a section type conflict
make[2]: *** [drivers/net/hp100.o] Error 1


It's really, as has been said, a compiler problem: the compiler is
confused about the read only sections.  However, 4.2 is a pretty common
compiler (especially for non-x86), so if the compiler can't do the read
only section tracking, just not using the __devinitconst designation
globally seems to be the best thing, just change it to __devinitdata
instead.

James

Re: Section conflict compile failures in net

[David Miller]

From: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Date: Wed, 1 Jun 2011 03:27:40 +0200

> __devinitconst ID tables do work for PCI drivers. What's different for EISA,
> and why is it working in x86 allyesconfig build but not in hppa?

James shows that it doesn't work for PCI drivers either.

Re: Section conflict compile failures in net

[David Miller]

From: James Bottomley <James.Bottomley@HansenPartnership.com>
Date: Wed, 01 Jun 2011 10:42:50 +0900

> It's really, as has been said, a compiler problem: the compiler is
> confused about the read only sections.  However, 4.2 is a pretty common
> compiler (especially for non-x86), so if the compiler can't do the read
> only section tracking, just not using the __devinitconst designation
> globally seems to be the best thing, just change it to __devinitdata
> instead.

I think I'm just going to do a full revert then.

You guys work out how to arrive at a solution that solves the problem
he was originally trying to solve, yet doesn't break this gcc-4.2
case.

Thanks.

Re: Section conflict compile failures in net

[Michał Mirosław]

On Wed, Jun 01, 2011 at 10:42:50AM +0900, James Bottomley wrote:
> On Tue, 2011-05-31 at 15:45 -0700, David Miller wrote:
> > From: James Bottomley <James.Bottomley@HansenPartnership.com>
> > Date: Thu, 26 May 2011 16:39:53 -0500
> > 
> > BTW, linux-netdev doesn't exist, it's just plain netdev.
> 
> Um, yes ... I thought it was netdev, then I checked MARC and it had
> linux-netdev.  I'll remember for next time (well, at least for the next
> six months).
> 
> > > Simply reverting 
> > > 
> > > commit e5cb966c0838e4da43a3b0751bdcac7fe719f7b4
> > > Author: Micha<C5><82> Miros<C5><82>aw <mirq-linux@rere.qmqm.pl>
> > > Date:   Mon Apr 18 13:31:20 2011 +0000
> > > 
> > >     net: fix section mismatches
> > >     
> > > Fixes the problem.
> > > 
> > > If I look at the first problem in hp100.c, the addition of
> > > __devinitconst to the device tables is fine, but there's no
> > > corresponding sectional tag on their use, so when compiled as a module,
> > > things like hp100_eisa_driver is now in the main data section but refers
> > > to something in the init data section ... I assume all the others are
> > > the same type of problem.
> > 
> > Yeah I think the device ID table __devinitconst bits have to be
> > reverted.
> > 
> > I'll apply something like the following:
> > 
> > net: Revert adding __devinitconst to driver ID tables.
> > 
> > If the table is hooked up to foo_driver->id_table we can't
> > mark it __devinitconst otherwise we end up with section
> > mismatches.
> > 
> > Reported-by: James Bottomley <James.Bottomley@HansenPartnership.com>
> > Signed-off-by: David S. Miller <davem@davemloft.net>
> 
> That fixes some but not all; the pci tables still cause a section
> conflict:
> 
> drivers/net/hp100.c:211: error: hp100_pci_tbl causes a section type conflict
> make[2]: *** [drivers/net/hp100.o] Error 1
> 
> 
> It's really, as has been said, a compiler problem: the compiler is
> confused about the read only sections.  However, 4.2 is a pretty common
> compiler (especially for non-x86), so if the compiler can't do the read
> only section tracking, just not using the __devinitconst designation
> globally seems to be the best thing, just change it to __devinitdata
> instead.

So maybe just #define __devinitconst __devinitdata for those broken gcc
versions?

Best Regards,
Michał Mirosław