Skip to contentSkip to conversation

2.4.23aa1

DISCUSSION Applied Andrea Arcangeli 3 participants 8 messages Dec 5, 2003 – Dec 9, 2003 I2C

lore
Share & tools
lore
#0 lore raw

This should be the last 2.4-aa kernel ;)

URL:

	http://www.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.23aa1.gz

Changelog diff between 2.4.22pre6aa3 and 2.4.23aa1:

Only in 2.4.23pre6aa3: 00_binfmt-elf-checks-3 Only in 2.4.23aa1: 00_binfmt-elf-checks-4 Only in 2.4.23pre6aa3: 00_extraversion-32 Only in 2.4.23aa1: 00_extraversion-33
Only in 2.4.23pre6aa3: 00_nfs_writeback-1 Only in 2.4.23aa1: 00_nfs_writeback-2
Only in 2.4.23pre6aa3: 00_silent-stack-overflow-19 Only in 2.4.23aa1: 00_silent-stack-overflow-20 Only in 2.4.23pre6aa3: 05_vm_22_vm-anon-lru-2 Only in 2.4.23aa1: 05_vm_22_vm-anon-lru-3 Only in 2.4.23pre6aa3: 05_vm_26-rest-1 Only in 2.4.23aa1: 05_vm_26-rest-2
Only in 2.4.23pre6aa3: 10_inode-highmem-2 Only in 2.4.23aa1: 10_inode-highmem-3
Only in 2.4.23pre6aa3: 60_tux-exports-7 Only in 2.4.23aa1: 60_tux-exports-8
Only in 2.4.23pre6aa3: 70_iget-1
Only in 2.4.23aa1: 70_iget-2
Only in 2.4.23pre6aa3: 93_NUMAQ-14
Only in 2.4.23aa1: 93_NUMAQ-16
Only in 2.4.23pre6aa3: 9900_aio-24.gz
Only in 2.4.23aa1: 9900_aio-25.gz
Only in 2.4.23pre6aa3: 9910_shm-largepage-17.gz Only in 2.4.23aa1: 9910_shm-largepage-18.gz Only in 2.4.23pre6aa3: 9920_kgdb-12.gz Only in 2.4.23aa1: 9920_kgdb-13.gz
Only in 2.4.23pre6aa3: 9925_kmsgdump-0.4.4-4.gz Only in 2.4.23aa1: 9925_kmsgdump-0.4.4-5.gz Only in 2.4.23pre6aa3: 9980_fix-pausing-6 Only in 2.4.23aa1: 9980_fix-pausing-7
Only in 2.4.23pre6aa3: 9985_blk-atomic-12 Only in 2.4.23aa1: 9985_blk-atomic-13
Only in 2.4.23pre6aa3: 9999900_q-full-1 Only in 2.4.23aa1: 9999900_q-full-2
Only in 2.4.23pre6aa3: 9999_mmap-cache-2 Only in 2.4.23aa1: 9999_mmap-cache-3
Only in 2.4.23pre6aa3: 9999_sched_yield_scale-8 Only in 2.4.23aa1: 9999_sched_yield_scale-9 Only in 2.4.23pre6aa3: 9999_zzz-dynamic-hz-3.gz Only in 2.4.23aa1: 9999_zzz-dynamic-hz-4.gz

Rediffed.

Only in 2.4.23pre6aa3: 00_do_brk-1
Only in 2.4.23pre6aa3: 00_pppoe-release-1
Only in 2.4.23pre6aa3: 00_proc-readlink-1
Only in 2.4.23pre6aa3: 00_sk98lin-char-fix-1
Only in 2.4.23pre6aa3: 00_tcp-conntrack-fin-1
Only in 2.4.23pre6aa3: 9999_01_x86_64-fault32-wrap-1 Only in 2.4.23pre6aa3: 9999900_shmem_truncate-against-swapoff-1 Only in 2.4.23pre6aa3: 9999_z-laptopmode-1

Merged in mainline.

Only in 2.4.23aa1: 00_readahead-last-page-1

Make sure to read the last page of a file via readahead if possible, otherwise we could send two synchronous requests down to the fs instead of a single one. Fix from Chuck Lever.

Only in 2.4.23aa1: 05_vm_00_anon-lru-race-better-fix-1

Serialize inserction in swapcache with the pagemap_lru_lock spinlock, to avoid having to use a flood of atomic ops (this benefits especially non x86 archs, for x86 if the cacheline is in l1 it's not a big cost). pagecache operations are unaffected this way.

Only in 2.4.23pre6aa3: 9999_z-execve-race-1 Only in 2.4.23aa1: 05_vm_02-execve-mm-fast-path-safe-1

	Avoid dropping the fast path in execve when mm_users == 1.
	Original fix from Ernie Petrides.

Only in 2.4.23aa1: 30_00_posix-race-1 Only in 2.4.23aa1: 30_02_fix-commit-1 Only in 2.4.23aa1: 30_03_fix-osx-1
Only in 2.4.23aa1: 30_04_soft2-1
Only in 2.4.23pre6aa3: 30_05_seekdir-1 Only in 2.4.23aa1: 30_05_seekdir-2
Only in 2.4.23pre6aa3: 30_08_rdplus-1 Only in 2.4.23aa1: 30_08_rdplus-2
Only in 2.4.23aa1: 30_09_fix-unlink-1 Only in 2.4.23aa1: 30_10-sock-disconnect-1 Only in 2.4.23pre6aa3: 30_12-lockd3-1 Only in 2.4.23aa1: 30_12-lockd3-2
Only in 2.4.23pre6aa3: 30_14-pathconf-2 Only in 2.4.23aa1: 30_14-pathconf-3
Only in 2.4.23pre6aa3: 30_18-busy-inodes-1

NFS updates from Trond's website.

Only in 2.4.23pre6aa3: 50_uml-patch-2.4.20-5-2.gz Only in 2.4.23aa1: 50_uml-patch-2.4.20-5-3.gz

Rediffed. The new version pretents to change the API for munmap in the f_ops, that's too invasive for 2.4 IMHO.

Only in 2.4.23pre6aa3: 90_ext3-commit-interval-5

Obsoleted by the laptop mode (I hope it's "fully" obsoleted ;).

Only in 2.4.23pre6aa3: 9986_elevator-merge-fast-path-2

	Dropped, seems to have a race of some sort (triggers on some
	ia64).
#1 lore raw
Quoted 4 lines · Andrea Arcangeli

Only in 2.4.23pre6aa3: 90_ext3-commit-interval-5

Obsoleted by the laptop mode (I hope it's "fully" obsoleted ;).

Maybe not. I saw one report here on lkml that showed that ext3-commit-interval is still needed even with laptop mode. :(

#2 lore raw
Quoted 8 lines · Mike Fedyk
Quoted 4 lines · Andrea Arcangeli

Only in 2.4.23pre6aa3: 90_ext3-commit-interval-5

Obsoleted by the laptop mode (I hope it's "fully" obsoleted ;).

Maybe not. I saw one report here on lkml that showed that ext3-commit-interval is still needed even with laptop mode. :(

It would be a bug of the laptop mode patch merged in mainline in such case. I tracked related changes in ext3, so I assumed it was just covered. If it isn't covered, laptop mode will be useless because the hd will never go to sleep ;)

#4 lore raw

Re: 2.4.23aa1 - scsi/pcmcia qlogic still does not build (m)

Quoted 7 lines · Eyal Lebedinsky
Quoted 2 lines · Andrea Arcangeli

This should be the last 2.4-aa kernel ;)

I guess nobody volunteered to fix it since -pre6aa2...

It builds just fine in vanilla 2.4.23.

is the error still the same as in your email with ID [email protected] right?

If nobody sends a fix I'll have another quick look after I'm back from vacations on Thu.

Also for the i2c troubles (you mentioned those last time), you can try if this helps.

drivers/i2c/i2c-elektor.c +1 -1
diff -urNp linux-2.4.21/drivers/i2c/i2c-elektor.c linux-2.4.21.SUSE/drivers/i2c/i2c-elektor.c
@@ -167,7 +167,6 @@ static struct i2c_algo_pcf_data pcf_isa_
 	.waitforpin = pcf_isa_waitforpin,
 	.udelay	    = 10,
 	.mdelay	    = 10,
-	.timeout    = HZ,
 };
 
 static struct i2c_adapter pcf_isa_ops = {
@@ -179,6 +178,7 @@ static struct i2c_adapter pcf_isa_ops = 
 
 static int __init i2c_pcfisa_init(void) 
 {
+	pcf_isa_data.timeout = HZ;
 #ifdef __alpha__
 	/* check to see we have memory mapped PCF8584 connected to the 
 	Cypress cy82c693 PCI-ISA bridge as on UP2000 board */
drivers/i2c/i2c-elv.c +1 -1
diff -urNp linux-2.4.21/drivers/i2c/i2c-elv.c linux-2.4.21.SUSE/drivers/i2c/i2c-elv.c
@@ -124,7 +124,6 @@ static struct i2c_algo_bit_data bit_elv_
 	.getscl		= bit_elv_getscl,
 	.udelay		= 80,
 	.mdelay		= 80,
-	.timeout	= HZ
 };
 
 static struct i2c_adapter bit_elv_ops = {
@@ -137,6 +136,7 @@ static struct i2c_adapter bit_elv_ops = 
 static int __init i2c_bitelv_init(void)
 {
 	printk(KERN_INFO "i2c-elv.o: i2c ELV parallel port adapter module version %s (%s)\n", I2C_VERSION, I2C_DATE);
+	bit_elv_data.timeout = HZ;
 	if (base==0) {
 		/* probe some values */
 		base=DEFAULT_BASE;
drivers/i2c/i2c-hydra.c +1 -1
diff -urNp linux-2.4.21/drivers/i2c/i2c-hydra.c linux-2.4.21.SUSE/drivers/i2c/i2c-hydra.c
@@ -107,7 +107,6 @@ static struct i2c_algo_bit_data bit_hydr
 	.getscl		= bit_hydra_getscl,
 	.udelay		= 5,
 	.mdelay		= 5,
-	.timeout	= HZ
 };
 
 static struct i2c_adapter bit_hydra_ops = {
@@ -154,6 +153,7 @@ static struct pci_driver hydra_driver = 
 
 static int __init i2c_hydra_init(void)
 {
+	bit_hydra_data.timeout = HZ;
 	return pci_module_init(&hydra_driver);
 }
 
drivers/i2c/i2c-i810.c +4 -2
diff -urNp linux-2.4.21/drivers/i2c/i2c-i810.c linux-2.4.21.SUSE/drivers/i2c/i2c-i810.c
@@ -189,7 +189,6 @@ static struct i2c_algo_bit_data i810_i2c
 	.getscl		= bit_i810i2c_getscl,
 	.udelay		= CYCLE_DELAY,
 	.mdelay		= CYCLE_DELAY,
-	.timeout	= TIMEOUT,
 };
 
 static struct i2c_adapter i810_i2c_adapter = {
@@ -206,7 +205,6 @@ static struct i2c_algo_bit_data i810_ddc
 	.getscl		= bit_i810ddc_getscl,
 	.udelay		= CYCLE_DELAY,
 	.mdelay		= CYCLE_DELAY,
-	.timeout	= TIMEOUT,
 };
 
 static struct i2c_adapter i810_ddc_adapter = {
@@ -293,6 +291,10 @@ static int __init i2c_i810_init(void)
 /*
 	return pci_module_init(&i810_driver);
 */
+
+	i810_i2c_bit_data.timeout = TIMEOUT;
+	i810_ddc_bit_data.timeout = TIMEOUT;
+
 	pci_for_each_dev(dev) {
 		id = pci_match_device(i810_ids, dev);
 		if(id)
drivers/i2c/i2c-philips-par.c +3 -2
diff -urNp linux-2.4.21/drivers/i2c/i2c-philips-par.c linux-2.4.21.SUSE/drivers/i2c/i2c-philips-par.c
@@ -137,7 +137,6 @@ static struct i2c_algo_bit_data bit_lp_d
 	.getscl		= bit_lp_getscl,
 	.udelay		= 80,
 	.mdelay		= 80,
-	.timeout	= HZ
 }; 
 
 static struct i2c_algo_bit_data bit_lp_data2 = {
@@ -146,7 +145,6 @@ static struct i2c_algo_bit_data bit_lp_d
 	.getsda		= bit_lp_getsda2,
 	.udelay		= 80,
 	.mdelay		= 80,
-	.timeout	= HZ
 }; 
 
 static struct i2c_adapter bit_lp_ops = {
@@ -236,6 +234,9 @@ int __init i2c_bitlp_init(void)
 {
 	printk(KERN_INFO "i2c-philips-par.o: i2c Philips parallel port adapter module version %s (%s)\n", I2C_VERSION, I2C_DATE);
 
+	bit_lp_data.timeout = HZ;
+	bit_lp_data2.timeout = HZ;
+
 	parport_register_driver(&i2c_driver);
 	
 	return 0;
drivers/i2c/i2c-savage4.c +3 -1
diff -urNp linux-2.4.21/drivers/i2c/i2c-savage4.c linux-2.4.21.SUSE/drivers/i2c/i2c-savage4.c
@@ -158,7 +158,6 @@ static struct i2c_algo_bit_data sav_i2c_
 	.getscl		= bit_savi2c_getscl,
 	.udelay		= CYCLE_DELAY,
 	.mdelay		= CYCLE_DELAY,
-	.timeout	= TIMEOUT
 };
 
 static struct i2c_adapter savage4_i2c_adapter = {
@@ -215,6 +214,9 @@ static int __init i2c_savage4_init(void)
 /*
 	return pci_module_init(&savage4_driver);
 */
+
+	sav_i2c_bit_data.timeout = TIMEOUT;
+
 	pci_for_each_dev(dev) {
 		id = pci_match_device(savage4_ids, dev);
 		if(id)
drivers/i2c/i2c-velleman.c +1 -1
diff -urNp linux-2.4.21/drivers/i2c/i2c-velleman.c linux-2.4.21.SUSE/drivers/i2c/i2c-velleman.c
@@ -110,7 +110,6 @@ static struct i2c_algo_bit_data bit_vell
 	.getscl		= bit_velle_getscl,
 	.udelay		= 10,
 	.mdelay		= 10,
-	.timeout	= HZ
 };
 
 static struct i2c_adapter bit_velle_ops = {
@@ -123,6 +122,7 @@ static struct i2c_adapter bit_velle_ops 
 static int __init i2c_bitvelle_init(void)
 {
 	printk(KERN_INFO "i2c-velleman.o: i2c Velleman K8000 adapter module version %s (%s)\n", I2C_VERSION, I2C_DATE);
+	bit_velle_data.timeout = HZ;
 	if (base==0) {
 		/* probe some values */
 		base=DEFAULT_BASE;
drivers/i2c/i2c-via.c +3 -1
diff -urNp linux-2.4.21/drivers/i2c/i2c-via.c linux-2.4.21.SUSE/drivers/i2c/i2c-via.c
@@ -89,7 +89,6 @@ static struct i2c_algo_bit_data bit_data
 	.getscl		= bit_via_getscl,
 	.udelay		= 5,
 	.mdelay		= 5,
-	.timeout	= HZ
 };
 
 static struct i2c_adapter vt586b_adapter = {
@@ -179,6 +178,9 @@ static int __init i2c_vt586b_init(void)
 /*
 	return pci_module_init(&vt586b_driver);
 */
+
+	bit_data.timeout = HZ;
+
 	pci_for_each_dev(dev) {
 		id = pci_match_device(vt586b_ids, dev);
 		if(id)
drivers/i2c/i2c-voodoo3.c +4 -2
diff -urNp linux-2.4.21/drivers/i2c/i2c-voodoo3.c linux-2.4.21.SUSE/drivers/i2c/i2c-voodoo3.c
@@ -178,7 +178,6 @@ static struct i2c_algo_bit_data voo_i2c_
 	.getscl		= bit_vooi2c_getscl,
 	.udelay		= CYCLE_DELAY,
 	.mdelay		= CYCLE_DELAY,
-	.timeout	= TIMEOUT
 };
 
 static struct i2c_adapter voodoo3_i2c_adapter = {
@@ -195,7 +194,6 @@ static struct i2c_algo_bit_data voo_ddc_
 	.getscl		= bit_vooddc_getscl,
 	.udelay		= CYCLE_DELAY,
 	.mdelay		= CYCLE_DELAY,
-	.timeout	= TIMEOUT
 };
 
 static struct i2c_adapter voodoo3_ddc_adapter = {
@@ -263,6 +261,10 @@ static int __init i2c_voodoo3_init(void)
 /*
 	return pci_module_init(&voodoo3_driver);
 */
+
+	voo_i2c_bit_data.timeout = TIMEOUT;
+	voo_ddc_bit_data.timeout = TIMEOUT;
+
 	pci_for_each_dev(dev) {
 		id = pci_match_device(voodoo3_ids, dev);
 		if(id)

Thanks.

#5 lore raw

Re: 2.4.23aa1 - scsi/pcmcia qlogic still does not build (m)

Quoted 4 lines · Andrea Arcangeli
Quoted 2 lines · Eyal Lebedinsky
Quoted 1 line · Andrea Arcangeli
Quoted 8 lines
Quoted 5 lines
Quoted 1 line

This should be the last 2.4-aa kernel ;)

I guess nobody volunteered to fix it since -pre6aa2...

It builds just fine in vanilla 2.4.23.

is the error still the same as in your email with ID [email protected] right?

Yes, problem with module_init. Same with fdomain etc..

I disabled the subsystem in order to complete the build.

Here is an example of the failure (for another module in this subsystem):

ld -m elf_i386 -r -o aha152x_cs.o aha152x_stub.o aha152x.o aha152x.o: In function `init_module':
aha152x.o(.text+0x50f0): multiple definition of `init_module' aha152x_stub.o(.text+0x740): first defined here
ld: Warning: size of symbol `init_module' changed from 77 to 58 in aha152x.o
aha152x.o: In function `cleanup_module':
aha152x.o(.text+0x5130): multiple definition of `cleanup_module' aha152x_stub.o(.text+0x790): first defined here
ld: Warning: size of symbol `cleanup_module' changed from 40 to 16 in aha152x.o
make[3]: *** [aha152x_cs.o] Error 1
make[3]: Leaving directory `/data2/usr/local/src/linux-2.4-aa/drivers/scsi/pcmcia'

--
Eyal Lebedinsky ([email protected]) <http://samba.org/eyal/>

#6 lore raw

Re: 2.4.23aa1 - scsi/pcmcia qlogic still does not build (m)

In reply to Andrea Arcangeli
Quoted 2 lines · Andrea Arcangeli

Also for the i2c troubles (you mentioned those last time), you can try if this helps.

I applied fixes similar to your suggestion to i2c-2.7.0 and lm_sensors-2.7.0 to get these to build, looks OK.

My final depmod has this problem though:

depmod: *** Unresolved symbols in /lib/modules/2.4.23aa1/kernel/drivers/video/sis/sisfb.o

depmod:         __floatsidf
depmod:         __divdf3
depmod:         __fixunsdfsi
depmod:         __muldf3
depmod:         __adddf3

I do not have this problem with 2.4.23, and I see the -aa1 patch actually removing some FP ops. But therer are some left in other sources (e.g. sis_main.c) so maybe some link problem was exposed?

Reverting the sis/init.c hunk does not fix this. Can it be related to this in arch/i386/Makefile:

-CFLAGS += -pipe
+CFLAGS += -pipe -msoft-float

--
Eyal Lebedinsky ([email protected]) <http://samba.org/eyal/>

#7 lore raw

Re: 2.4.23aa1 - scsi/pcmcia qlogic still does not build (m)

Quoted 17 lines · Eyal Lebedinsky
Quoted 2 lines · Andrea Arcangeli

Also for the i2c troubles (you mentioned those last time), you can try if this helps.

I applied fixes similar to your suggestion to i2c-2.7.0 and lm_sensors-2.7.0 to get these to build, looks OK.

My final depmod has this problem though:

depmod: *** Unresolved symbols in /lib/modules/2.4.23aa1/kernel/drivers/video/sis/sisfb.o

depmod:         __floatsidf
depmod:         __divdf3
depmod:         __fixunsdfsi
depmod:         __muldf3
depmod:         __adddf3
Quoted 3 lines

I do not have this problem with 2.4.23, and I see the -aa1 patch actually removing some FP ops. But therer are some left in other sources (e.g. sis_main.c) so maybe some link problem was exposed?

yes I've a 150k compressed updated driver from Thomas Winischhofer in my inbox for Marcelo that should fix those bugs (that would obsolete the non complete 00_sis-fpu-bugs-1), I thought it was merged in mainline but obviously not as 00_sis-fpu-bugs-1 wouldn't apply anymore. I guess Marcelo rejected it because it was very big and it wasn't fix the strict fpu bugs revealed by the -msoft-float, just guessing.

Review

Quoted 5 lines

Reverting the sis/init.c hunk does not fix this. Can it be related to this in arch/i386/Makefile:

(quoted diff) +1 -1
-CFLAGS += -pipe
+CFLAGS += -pipe -msoft-float

yes it's related, reverting it would hide the bug, the module would load again but userspace could be corrupted at runtime.