Skip to contentSkip to conversation

feature-removal-schedule obsoletes

DISCUSSION Jiri Slaby 8 participants 15 messages Oct 15, 2006 – Nov 13, 2006 MMC, SD and SDIO

lore
Share & tools
lore

Jiri Slaby

author
#0 lore raw

Hello,

I figured out these requests for removal with past dates. What's the state of them now, could you reschedule or delete (or just confirm here to post one patch correcting this file)?

[Why:s were removed in this mail]

---------------------------

What:   RAW driver (CONFIG_RAW_DRIVER)
When:   December 2005
Who:    Adrian Bunk <[email protected]>

---------------------------

What:   PCMCIA control ioctl (needed for pcmcia-cs [cardmgr, cardctl])
When:   November 2005
Who:    Dominik Brodowski <[email protected]>

---------------------------

What:   ip_queue and ip6_queue (old ipv4-only and ipv6-only netfilter queue)
When:   December 2005
Who:    Harald Welte <[email protected]>

---------------------------

What:   remove EXPORT_SYMBOL(kernel_thread)
When:   August 2006
Who:    Christoph Hellwig <[email protected]>

---------------------------

What:   CONFIG_FORCED_INLINING
When:   June 2006
Who:    Arjan van de Ven

---------------------------

What: Unused EXPORT_SYMBOL/EXPORT_SYMBOL_GPL exports
(temporary transition config option provided until then) The transition config option will also be removed at the same time.

When:   before 2.6.19
Who:    Arjan van de Ven <[email protected]>

---------------------------

What:   i2c-ite and i2c-algo-ite drivers
When:   September 2006
Who:    Jean Delvare <[email protected]>

---------------------------

thanks,

Signature · 4 lines
http://www.fi.muni.cz/~xslaby/            Jiri Slaby
faculty of informatics, masaryk university, brno, cz
e-mail: jirislaby gmail com, gpg pubkey fingerprint:
B674 9967 0407 CE62 ACC8  22A0 32CC 55C3 39D4 7A7E
#1 lore raw

Hi Jiri,

Quoted 8 lines · Jiri Slaby

I figured out these requests for removal with past dates. What's the state of them now, could you reschedule or delete (or just confirm here to post one patch correcting this file)?

[Why:s were removed in this mail]
(...)
---------------------------

Quoted 5 lines
What:   i2c-ite and i2c-algo-ite drivers
When:   September 2006
Who:    Jean Delvare <[email protected]>

---------------------------

The patch is ready, it's in my local stack:
http://khali.linux-fr.org/devel/linux-2.6/i2c-delete-ite-bus-driver.patch

I'll push it upstream soon.

Thanks,

Signature · 1 line
Jean Delvare
#3 lore raw
Quoted 7 lines · Christoph Hellwig
Quoted 3 lines · Jiri Slaby
What:   remove EXPORT_SYMBOL(kernel_thread)
When:   August 2006
Who:    Christoph Hellwig <[email protected]>

There are a lot of modular users left. It'll go away as soon as these users have disappeared.

It seems that most of the users that are left are for pretty obscure functionality, so I wouldn't expect that to happen so soon. Maybe we should mark it as __deprecated in the declaration?

	Arnd <><

./arch/arm/kernel/ecard.c: ret = kernel_thread(ecard_task, NULL, CLONE_KERNEL);
./arch/arm/kernel/process.c:pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
./arch/i386/kernel/io_apic.c: if (kernel_thread(balanced_irq, NULL, CLONE_KERNEL) >= 0)
./arch/i386/mach-voyager/voyager_thread.c: if(kernel_thread(thread, NULL, CLONE_KERNEL) < 0) {
./arch/ia64/sn/kernel/xpc_main.c: pid = kernel_thread(xpc_activating, (void *) ((u64) partid), 0);
./drivers/macintosh/adb.c: adb_probe_task_pid = kernel_thread(adb_probe_task, NULL, SIGCHLD | CLONE_KERNEL);
./drivers/macintosh/mediabay.c: kernel_thread(media_bay_task, NULL, CLONE_KERNEL);
./drivers/macintosh/therm_pm72.c: ctrl_task = kernel_thread(main_control_loop, NULL, SIGCHLD | CLONE_KERNEL);
./drivers/macintosh/therm_windtunnel.c: x.poll_task = kernel_thread( control_loop, NULL, SIGCHLD | CLONE_KERNEL );
./drivers/media/dvb/dvb-core/dvb_ca_en50221.c: ret = kernel_thread(dvb_ca_en50221_thread, ca, 0);
./drivers/media/dvb/dvb-core/dvb_frontend.c: ret = kernel_thread (dvb_frontend_thread, fe, 0);
./drivers/media/dvb/ttpci/av7110.c: ret = kernel_thread(arm_thread, (void *) av7110, 0);
./drivers/media/video/saa7134/saa7134-tvaudio.c: dev->thread.pid = kernel_thread(my_thread,dev,0);
./drivers/media/video/msp3400-driver.c: v4l_warn(client, "kernel_thread() failed\n");
./drivers/mmc/mmc_queue.c: ret = kernel_thread(mmc_queue_thread, mq, CLONE_KERNEL);
./drivers/mtd/mtd_blkdevs.c: ret = kernel_thread(mtd_blktrans_thread, tr, CLONE_KERNEL);
./drivers/pci/hotplug/cpci_hotplug_core.c: pid = kernel_thread(event_thread, NULL, 0);
./drivers/pci/hotplug/cpci_hotplug_core.c: pid = kernel_thread(poll_thread, NULL, 0);

./drivers/pci/hotplug/cpqphp_ctrl.c:    pid = kernel_thread(event_thread, NULL, 0);
./drivers/pci/hotplug/ibmphp_hpc.c:     tid_poll = kernel_thread (hpc_poll_thread, NULL, 0);
./drivers/pci/hotplug/pciehp_ctrl.c:    pid = kernel_thread(event_thread, NULL, 0);

./drivers/pnp/pnpbios/core.c: if (kernel_thread(pnp_dock_thread, NULL, CLONE_KERNEL) > 0)
./drivers/s390/net/lcs.c: kernel_thread(lcs_recovery, (void *) card, SIGCHLD);
./drivers/s390/net/qeth_main.c: kernel_thread(qeth_register_ip_addresses, (void *)card,SIGCHLD);
./drivers/s390/scsi/zfcp_erp.c: retval = kernel_thread(zfcp_erp_thread, adapter, SIGCHLD); ./drivers/scsi/libsas/sas_scsi_host.c: res = kernel_thread(sas_queue_thread, sas_ha, 0);
./drivers/usb/atm/usbatm.c: int ret = kernel_thread(usbatm_do_heavy_init, instance, CLONE_KERNEL);
./drivers/usb/atm/usbatm.c: usb_err(instance, "%s: failed to create kernel_thread (%d)!\n", __func__, ret);
./fs/afs/cmservice.c: ret = kernel_thread(kafscmd, NULL, 0);
./fs/afs/kafsasyncd.c: ret = kernel_thread(kafsasyncd, NULL, 0);
./fs/afs/kafstimod.c: ret = kernel_thread(kafstimod, NULL, 0);
./fs/cifs/connect.c: rc = (int)kernel_thread((void *)(void *)cifs_demultiplex_thread, srvTcp,
./fs/jffs/inode-v23.c: c->thread_pid = kernel_thread (jffs_garbage_collect_thread,
./fs/jffs2/background.c: pid = kernel_thread(jffs2_garbage_collect_thread, c, CLONE_FS|CLONE_FILES);
./fs/lockd/clntlock.c: if (kernel_thread(reclaimer, host, CLONE_KERNEL) < 0)
./fs/nfs/delegation.c: status = kernel_thread(recall_thread, &data, CLONE_KERNEL);

./net/bluetooth/bnep/core.c:    err = kernel_thread(bnep_session, s, CLONE_KERNEL);
./net/bluetooth/cmtp/core.c:    err = kernel_thread(cmtp_session, session, CLONE_KERNEL);
./net/bluetooth/hidp/core.c:    err = kernel_thread(hidp_session, session, CLONE_KERNEL);
./net/bluetooth/rfcomm/core.c:  kernel_thread(rfcomm_run, NULL, CLONE_KERNEL);
./net/core/pktgen.c:    err = kernel_thread((void *)pktgen_thread_worker, (void *)t,
./net/core/pktgen.c:            printk("pktgen: kernel_thread() failed for cpu %d\n", t->cpu);
./net/ipv4/ipvs/ip_vs_sync.c:   if ((pid = kernel_thread(sync_thread, startup, 0)) < 0) {
./net/ipv4/ipvs/ip_vs_sync.c:   if ((pid = kernel_thread(fork_sync_thread, &startup, 0)) < 0) {

./net/rxrpc/krxiod.c: return kernel_thread(rxrpc_krxiod, NULL, 0);
./net/rxrpc/krxsecd.c: return kernel_thread(rxrpc_krxsecd, NULL, 0); ./net/rxrpc/krxtimod.c: ret = kernel_thread(krxtimod, NULL, 0);
./net/sunrpc/svc.c: error = kernel_thread((int (*)(void *)) func, rqstp, 0);

#4 lore raw
Quoted 6 lines · Arnd Bergmann
Quoted 5 lines · Christoph Hellwig
Quoted 3 lines · Jiri Slaby
What:   remove EXPORT_SYMBOL(kernel_thread)
When:   August 2006
Who:    Christoph Hellwig <[email protected]>
Quoted 6 lines
Quoted 2 lines

There are a lot of modular users left. It'll go away as soon as these users have disappeared.

It seems that most of the users that are left are for pretty obscure functionality, so I wouldn't expect that to happen so soon. Maybe we should mark it as __deprecated in the declaration?

__deprecated_for_modules (__deprecated would imply it would be completely removed).

Quoted 1 line
	Arnd <><

cu
Adrian

Signature · 4 lines
       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed
#5 lore raw
In reply to Arnd Bergmann
Quoted 2 lines · Arnd Bergmann

./drivers/mtd/mtd_blkdevs.c: ret = kernel_thread(mtd_blktrans_thread, tr, CLONE_KERNEL);

This one should be ripped out. I did it as part of a bugfix for a customer kernel not too long ago.

Jörn

Signature · 2 lines
Joern's library part 3:
http://inst.eecs.berkeley.edu/~cs152/fa05/handouts/clark-test.pdf
#6 lore raw
In reply to Arnd Bergmann
Quoted 5 lines · Arnd Bergmann
Quoted 4 lines · Christoph Hellwig
Quoted 3 lines · Jiri Slaby
What:   remove EXPORT_SYMBOL(kernel_thread)
When:   August 2006
Who:    Christoph Hellwig <[email protected]>
Quoted 7 lines
Quoted 2 lines

There are a lot of modular users left. It'll go away as soon as these users have disappeared.

It seems that most of the users that are left are for pretty obscure functionality, so I wouldn't expect that to happen so soon. Maybe we should mark it as __deprecated in the declaration?

What should be used to replace it? The MMC block driver uses it to manage the block device queue. I am not that intimate with the block layer so I do not know the proper fix.

Rgds

Signature · 5 lines
     -- Pierre Ossman

  Linux kernel, MMC maintainer        http://www.kernel.org
  PulseAudio, core developer          http://pulseaudio.org
  rdesktop, core developer          http://www.rdesktop.org
#7 lore raw
Quoted 5 lines · Pierre Ossman

What should be used to replace it? The MMC block driver uses it to manage the block device queue. I am not that intimate with the block layer so I do not know the proper fix.

You should use kthread_create().

	Arnd <><
#8 lore raw
In reply to Pierre Ossman
Quoted 8 lines · Pierre Ossman
Quoted 4 lines

It seems that most of the users that are left are for pretty obscure functionality, so I wouldn't expect that to happen so soon. Maybe we should mark it as __deprecated in the declaration?

What should be used to replace it? The MMC block driver uses it to manage the block device queue. I am not that intimate with the block layer so I do not know the proper fix.

kthread_create/kthread_run. Here's a draft patch (and it's against a rather old tree and untested due to lack of hardware so it really should be considered just a draft).

drivers/mmc/mmc_queue.c +13 -26
Index: linux-2.6/drivers/mmc/mmc_queue.c
===================================================================
@@ -10,12 +10,12 @@
  */
 #include <linux/module.h>
 #include <linux/blkdev.h>
+#include <linux/kthread.h>
 
 #include <linux/mmc/card.h>
 #include <linux/mmc/host.h>
 #include "mmc_queue.h"
 
-#define MMC_QUEUE_EXIT		(1 << 0)
 #define MMC_QUEUE_SUSPENDED	(1 << 1)
 
 /*
@@ -59,7 +59,6 @@
 {
 	struct mmc_queue *mq = d;
 	struct request_queue *q = mq->queue;
-	DECLARE_WAITQUEUE(wait, current);
 
 	/*
 	 * Set iothread to ensure that we aren't put to sleep by
@@ -67,12 +66,7 @@
 	 */
 	current->flags |= PF_MEMALLOC|PF_NOFREEZE;
 
-	daemonize("mmcqd");
-
-	complete(&mq->thread_complete);
-
 	down(&mq->thread_sem);
-	add_wait_queue(&mq->thread_wq, &wait);
 	do {
 		struct request *req = NULL;
 
@@ -84,7 +78,7 @@
 		spin_unlock_irq(q->queue_lock);
 
 		if (!req) {
-			if (mq->flags & MMC_QUEUE_EXIT)
+			if (kthread_should_stop())
 				break;
 			up(&mq->thread_sem);
 			schedule();
@@ -95,10 +89,8 @@
 
 		mq->issue_fn(mq, req);
 	} while (1);
-	remove_wait_queue(&mq->thread_wq, &wait);
 	up(&mq->thread_sem);
 
-	complete_and_exit(&mq->thread_complete, 0);
 	return 0;
 }
 
@@ -113,7 +105,7 @@
 	struct mmc_queue *mq = q->queuedata;
 
 	if (!mq->req)
-		wake_up(&mq->thread_wq);
+		wake_up_process(mq->thread);
 }
 
 /**
@@ -152,36 +144,31 @@
 			 GFP_KERNEL);
 	if (!mq->sg) {
 		ret = -ENOMEM;
-		goto cleanup;
+		goto cleanup_queue;
 	}
 
-	init_completion(&mq->thread_complete);
-	init_waitqueue_head(&mq->thread_wq);
 	init_MUTEX(&mq->thread_sem);
 
-	ret = kernel_thread(mmc_queue_thread, mq, CLONE_KERNEL);
-	if (ret >= 0) {
-		wait_for_completion(&mq->thread_complete);
-		init_completion(&mq->thread_complete);
-		ret = 0;
-		goto out;
+	mq->thread = kthread_run(mmc_queue_thread, mq, "mmcqd");
+	if (IS_ERR(mq->thread)) {
+		ret = PTR_ERR(mq->thread);
+		goto free_sg;
 	}
 
- cleanup:
+	return 0;
+
+ free_sg:
 	kfree(mq->sg);
 	mq->sg = NULL;
-
+ cleanup_queue:
 	blk_cleanup_queue(mq->queue);
- out:
 	return ret;
 }
 EXPORT_SYMBOL(mmc_init_queue);
 
 void mmc_cleanup_queue(struct mmc_queue *mq)
 {
-	mq->flags |= MMC_QUEUE_EXIT;
-	wake_up(&mq->thread_wq);
-	wait_for_completion(&mq->thread_complete);
+	kthread_stop(mq->thread);
 
 	kfree(mq->sg);
 	mq->sg = NULL;
drivers/mmc/mmc_queue.h +1 -2
Index: linux-2.6/drivers/mmc/mmc_queue.h
===================================================================
@@ -6,8 +6,7 @@
 
 struct mmc_queue {
 	struct mmc_card		*card;
-	struct completion	thread_complete;
-	wait_queue_head_t	thread_wq;
+	struct task_struct	*thread;
 	struct semaphore	thread_sem;
 	unsigned int		flags;
 	struct request		*req;
#9 lore raw
Quoted 12 lines · Christoph Hellwig
Quoted 4 lines · Pierre Ossman

What should be used to replace it? The MMC block driver uses it to manage the block device queue. I am not that intimate with the block layer so I do not know the proper fix.

kthread_create/kthread_run. Here's a draft patch (and it's against a rather old tree and untested due to lack of hardware so it really should be considered just a draft).

Patch looks ok to me and seems to work fine. Are you willing to sign off on it?

Rgds

Signature · 5 lines
     -- Pierre Ossman

  Linux kernel, MMC maintainer        http://www.kernel.org
  PulseAudio, core developer          http://pulseaudio.org
  rdesktop, core developer          http://www.rdesktop.org
#10 lore raw
Quoted 8 lines · Pierre Ossman
Quoted 7 lines · Christoph Hellwig
Quoted 4 lines · Pierre Ossman

What should be used to replace it? The MMC block driver uses it to manage the block device queue. I am not that intimate with the block layer so I do not know the proper fix.

Quoted 8 lines
Quoted 5 lines

kthread_create/kthread_run. Here's a draft patch (and it's against a rather old tree and untested due to lack of hardware so it really should be considered just a draft).

Patch looks ok to me and seems to work fine. Are you willing to sign off on it?

Sure, please add a

to the patch.

#11 lore raw
In reply to Pierre Ossman
Quoted 4 lines · Pierre Ossman

What should be used to replace it? The MMC block driver uses it to manage the block device queue. I am not that intimate with the block layer so I do not know the proper fix.

Why does the MMC block driver use a thread? Is there a technical reason for this or could it be done in original process context as well, removing some code and useless cpu scheduler overhead?

Jörn

Signature · 3 lines
When people work hard for you for a pat on the back, you've got
to give them that pat.
-- Robert Heinlein
#12 lore raw
Quoted 9 lines · J?rn Engel
Quoted 4 lines · Pierre Ossman

What should be used to replace it? The MMC block driver uses it to manage the block device queue. I am not that intimate with the block layer so I do not know the proper fix.

Why does the MMC block driver use a thread? Is there a technical reason for this or could it be done in original process context as well, removing some code and useless cpu scheduler overhead?

As I understand it, there is no guarantee that a block drivers request function will be called in process context - it could be called in interrupt context.

The MMC subsystem needs process context to issue commands since the process of issuing commands entails various sleeps. Hence why the MMC block has its own process context.

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core
#13 lore raw
Quoted 9 lines · Russell King
Quoted 4 lines · J?rn Engel

Why does the MMC block driver use a thread? Is there a technical reason for this or could it be done in original process context as well, removing some code and useless cpu scheduler overhead?

As I understand it, there is no guarantee that a block drivers request function will be called in process context - it could be called in interrupt context.

Makes some sense. I would still like to understand when a request function is called from interrupt context, but if it is, the thread is certainly necessary.

Jörn

Signature · 2 lines
This above all: to thine own self be true.
-- Shakespeare
#14 lore raw
In reply to Jörn Engel
Quoted 4 lines · Jörn Engel

Why does the MMC block driver use a thread? Is there a technical reason for this or could it be done in original process context as well, removing some code and useless cpu scheduler overhead?

I'm afraid I don't know the block layer very well, but that thread seems to be polling the block layer for requests and handing the over to the routines in mmc_block.c.

How do you set it up so that the block layer itself calls the necessary function?

Rgds

Signature · 5 lines
     -- Pierre Ossman

  Linux kernel, MMC maintainer        http://www.kernel.org
  PulseAudio, core developer          http://pulseaudio.org
  rdesktop, core developer          http://www.rdesktop.org