Skip to contentSkip to conversation

decruft the early init / initrd / initramfs code v2

PATCH In review Christoph Hellwig 5 participants 26 messages Jul 9, 2020 – Jul 14, 2020

Subsystems (8)
Arm (32-bit) x86 VFS SuperH +4 more subsystems
lore
Share & tools
lore
#0 lore raw

Hi all,

this series starts to move the early init code away from requiring KERNEL_DS to be implicitly set during early startup. It does so by first removing legacy unused cruft, and the switches away the code from struct file based APIs to our more usual in-kernel APIs.

There is no really good tree for this, so if there are no objections I'd like to set up a new one for linux-next.

Git tree:

git://git.infradead.org/users/hch/misc.git init-user-pointers

Gitweb:

http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/init-user-pointers

Changes since v1:
- add a patch to deprecated "classic" initrd support

Diffstat:

 b/arch/arm/kernel/atags_parse.c |    2 
 b/arch/sh/kernel/setup.c        |    2 
 b/arch/sparc/kernel/setup_32.c  |    2 
 b/arch/sparc/kernel/setup_64.c  |    2 
 b/arch/x86/kernel/setup.c       |    2 
 b/drivers/md/Makefile           |    3 
 b/drivers/md/md-autodetect.c    |  239 ++++++++++++++++++----------------------
 b/drivers/md/md.c               |   34 +----
 b/drivers/md/md.h               |   10 +
 b/fs/file.c                     |    7 -
 b/fs/open.c                     |   18 +--
 b/fs/read_write.c               |    2 
 b/fs/readdir.c                  |   11 -
 b/include/linux/initrd.h        |    6 -
 b/include/linux/raid/detect.h   |    8 +
 b/include/linux/syscalls.h      |   16 --
 b/init/Makefile                 |    1 
 b/init/do_mounts.c              |   70 +----------
 b/init/do_mounts.h              |   21 ---
 b/init/do_mounts_initrd.c       |   13 --
 b/init/do_mounts_rd.c           |  102 +++++++----------
 b/init/initramfs.c              |  103 +++++------------
 b/init/main.c                   |   16 +-
 include/linux/raid/md_u.h       |   13 --
 24 files changed, 251 insertions(+), 452 deletions(-)
#1 lore raw

PATCH 1/17 init: remove the bstat helper

The only caller of the bstat function becomes cleaner and simpler when open coding the function.

---
 init/do_mounts.h    | 10 ----------
 init/do_mounts_md.c |  8 ++++----
 2 files changed, 4 insertions(+), 14 deletions(-)
init/do_mounts.h +0 -10
index 0bb0806de4ce2c..7513d1c14d13fe 100644
@@ -20,16 +20,6 @@ static inline int create_dev(char *name, dev_t dev)
 	return ksys_mknod(name, S_IFBLK|0600, new_encode_dev(dev));
 }
 
-static inline u32 bstat(char *name)
-{
-	struct kstat stat;
-	if (vfs_stat(name, &stat) != 0)
-		return 0;
-	if (!S_ISBLK(stat.mode))
-		return 0;
-	return stat.rdev;
-}
-
 #ifdef CONFIG_BLK_DEV_RAM
 
 int __init rd_load_disk(int n);
init/do_mounts_md.c +4 -4
index b84031528dd446..359363e85ccd0b 100644
@@ -138,9 +138,9 @@ static void __init md_setup_drive(void)
 			dev = MKDEV(MD_MAJOR, minor);
 		create_dev(name, dev);
 		for (i = 0; i < MD_SB_DISKS && devname != NULL; i++) {
+			struct kstat stat;
 			char *p;
 			char comp_name[64];
-			u32 rdev;
 
 			p = strchr(devname, ',');
 			if (p)
@@ -150,9 +150,9 @@ static void __init md_setup_drive(void)
 			if (strncmp(devname, "/dev/", 5) == 0)
 				devname += 5;
 			snprintf(comp_name, 63, "/dev/%s", devname);
-			rdev = bstat(comp_name);
-			if (rdev)
-				dev = new_decode_dev(rdev);
+			if (vfs_stat(comp_name, &stat) == 0 &&
+			    S_ISBLK(stat.mode))
+				dev = new_decode_dev(stat.rdev);
 			if (!dev) {
 				printk(KERN_WARNING "md: Unknown device name: %s\n", devname);
 				break;
Signature · 1 line
2.26.2
#2 lore raw

PATCH 2/17 md: move the early init autodetect code to drivers/md/

In reply to Christoph Hellwig

Just like the NFS and CIFS root code this better lives with the driver it is tightly integrated with.

---
 drivers/md/Makefile                               |  3 +++
 init/do_mounts_md.c => drivers/md/md-autodetect.c | 15 +++++++++++++--
 include/linux/raid/detect.h                       |  8 ++++++++
 init/Makefile                                     |  1 -
 init/do_mounts.c                                  |  1 +
 init/do_mounts.h                                  | 10 ----------
 6 files changed, 25 insertions(+), 13 deletions(-)
 rename init/do_mounts_md.c => drivers/md/md-autodetect.c (96%)
drivers/md/Makefile +3 -0
index 31840f95cd408b..6d3e234dc46a5d 100644
@@ -43,6 +43,9 @@ obj-$(CONFIG_MD_FAULTY)		+= faulty.o
 obj-$(CONFIG_MD_CLUSTER)	+= md-cluster.o
 obj-$(CONFIG_BCACHE)		+= bcache/
 obj-$(CONFIG_BLK_DEV_MD)	+= md-mod.o
+ifeq ($(CONFIG_BLK_DEV_MD),y)
+obj-y				+= md-autodetect.o
+endif
 obj-$(CONFIG_BLK_DEV_DM)	+= dm-mod.o
 obj-$(CONFIG_BLK_DEV_DM_BUILTIN) += dm-builtin.o
 obj-$(CONFIG_DM_UNSTRIPED)	+= dm-unstripe.o
drivers/md/md-autodetect.c renamed from init/do_mounts_md.c +13 -2
similarity index 96%
rename from init/do_mounts_md.c
rename to drivers/md/md-autodetect.c
index 359363e85ccd0b..fe806f7b9759a1 100644
@@ -1,10 +1,15 @@
 // SPDX-License-Identifier: GPL-2.0
+#include <linux/kernel.h>
+#include <linux/blkdev.h>
+#include <linux/init.h>
+#include <linux/syscalls.h>
+#include <linux/mount.h>
+#include <linux/major.h>
 #include <linux/delay.h>
+#include <linux/raid/detect.h>
 #include <linux/raid/md_u.h>
 #include <linux/raid/md_p.h>
 
-#include "do_mounts.h"
-
 /*
  * When md (and any require personalities) are compiled into the kernel
  * (not a module), arrays can be assembles are boot time using with AUTODETECT
@@ -114,6 +119,12 @@ static int __init md_setup(char *str)
 	return 1;
 }
 
+static inline int create_dev(char *name, dev_t dev)
+{
+	ksys_unlink(name);
+	return ksys_mknod(name, S_IFBLK|0600, new_encode_dev(dev));
+}
+
 static void __init md_setup_drive(void)
 {
 	int minor, i, ent, partitioned;
include/linux/raid/detect.h +8 -0
index 37dd3f40cd316e..1f029a71c3ef05 100644
@@ -1,3 +1,11 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 
 void md_autodetect_dev(dev_t dev);
+
+#ifdef CONFIG_BLK_DEV_MD
+void md_run_setup(void);
+#else
+static inline void md_run_setup(void)
+{
+}
+#endif
init/Makefile +0 -1
index 57499b1ff4714d..6bc37f64b3617c 100644
@@ -18,7 +18,6 @@ obj-y                          += init_task.o
 mounts-y			:= do_mounts.o
 mounts-$(CONFIG_BLK_DEV_RAM)	+= do_mounts_rd.o
 mounts-$(CONFIG_BLK_DEV_INITRD)	+= do_mounts_initrd.o
-mounts-$(CONFIG_BLK_DEV_MD)	+= do_mounts_md.o
 
 # dependencies on generated files need to be listed explicitly
 $(obj)/version.o: include/generated/compile.h
init/do_mounts.c +1 -0
index 29d326b6c29d2d..1a4dfa17fb2899 100644
@@ -23,6 +23,7 @@
 #include <linux/nfs_fs.h>
 #include <linux/nfs_fs_sb.h>
 #include <linux/nfs_mount.h>
+#include <linux/raid/detect.h>
 #include <uapi/linux/mount.h>
 
 #include "do_mounts.h"
init/do_mounts.h +0 -10
index 7513d1c14d13fe..50d6c8941e15a1 100644
@@ -41,13 +41,3 @@ bool __init initrd_load(void);
 static inline bool initrd_load(void) { return false; }
 
 #endif
-
-#ifdef CONFIG_BLK_DEV_MD
-
-void md_run_setup(void);
-
-#else
-
-static inline void md_run_setup(void) {}
-
-#endif
Signature · 1 line
2.26.2
#3 lore raw

PATCH 3/17 md: replace the RAID_AUTORUN ioctl with a direct function call

In reply to Christoph Hellwig

Instead of using a spcial RAID_AUTORUN ioctl that only exists for non-modular builds and is only called from the early init code, just call the actual function directly.

---
 drivers/md/md-autodetect.c | 10 ++--------
 drivers/md/md.c            | 14 +-------------
 drivers/md/md.h            |  3 +++
 3 files changed, 6 insertions(+), 21 deletions(-)
drivers/md/md-autodetect.c +2 -8
index fe806f7b9759a1..0eb746211ed53c 100644
@@ -9,6 +9,7 @@
 #include <linux/raid/detect.h>
 #include <linux/raid/md_u.h>
 #include <linux/raid/md_p.h>
+#include "md.h"
 
 /*
  * When md (and any require personalities) are compiled into the kernel
@@ -285,8 +286,6 @@ __setup("md=", md_setup);
 
 static void __init autodetect_raid(void)
 {
-	int fd;
-
 	/*
 	 * Since we don't want to detect and use half a raid array, we need to
 	 * wait for the known devices to complete their probing
@@ -295,12 +294,7 @@ static void __init autodetect_raid(void)
 	printk(KERN_INFO "md: If you don't use raid, use raid=noautodetect\n");
 
 	wait_for_device_probe();
-
-	fd = ksys_open("/dev/md0", 0, 0);
-	if (fd >= 0) {
-		ksys_ioctl(fd, RAID_AUTORUN, raid_autopart);
-		ksys_close(fd);
-	}
+	md_autostart_arrays(raid_autopart);
 }
 
 void __init md_run_setup(void)
drivers/md/md.c +1 -13
index f567f536b529bd..6e9a48da474848 100644
@@ -68,10 +68,6 @@
 #include "md-bitmap.h"
 #include "md-cluster.h"
 
-#ifndef MODULE
-static void autostart_arrays(int part);
-#endif
-
 /* pers_list is a list of registered personalities protected
  * by pers_lock.
  * pers_lock does extra service to protect accesses to
@@ -7421,7 +7417,6 @@ static inline bool md_ioctl_valid(unsigned int cmd)
 	case GET_DISK_INFO:
 	case HOT_ADD_DISK:
 	case HOT_REMOVE_DISK:
-	case RAID_AUTORUN:
 	case RAID_VERSION:
 	case RESTART_ARRAY_RW:
 	case RUN_ARRAY:
@@ -7467,13 +7462,6 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 	case RAID_VERSION:
 		err = get_version(argp);
 		goto out;
-
-#ifndef MODULE
-	case RAID_AUTORUN:
-		err = 0;
-		autostart_arrays(arg);
-		goto out;
-#endif
 	default:;
 	}
 
@@ -9721,7 +9709,7 @@ void md_autodetect_dev(dev_t dev)
 	}
 }
 
-static void autostart_arrays(int part)
+void md_autostart_arrays(int part)
 {
 	struct md_rdev *rdev;
 	struct detected_devices_node *node_detected_dev;
drivers/md/md.h +3 -0
index 612814d07d35ab..37315a3f28e97d 100644
@@ -800,4 +800,7 @@ static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio
 	    !bio->bi_disk->queue->limits.max_write_zeroes_sectors)
 		mddev->queue->limits.max_write_zeroes_sectors = 0;
 }
+
+void md_autostart_arrays(int part);
+
 #endif /* _MD_MD_H */
Signature · 1 line
2.26.2
#4 lore raw

PATCH 4/17 md: remove the autoscan partition re-read

In reply to Christoph Hellwig

devfs is long gone, and autoscan works just fine without this these days.

---
 drivers/md/md-autodetect.c | 10 ----------
 1 file changed, 10 deletions(-)
drivers/md/md-autodetect.c +0 -10
index 0eb746211ed53c..6bc9b734eee6ff 100644
@@ -240,16 +240,6 @@ static void __init md_setup_drive(void)
 			err = ksys_ioctl(fd, RUN_ARRAY, 0);
 		if (err)
 			printk(KERN_WARNING "md: starting md%d failed\n", minor);
-		else {
-			/* reread the partition table.
-			 * I (neilb) and not sure why this is needed, but I cannot
-			 * boot a kernel with devfs compiled in from partitioned md
-			 * array without it
-			 */
-			ksys_close(fd);
-			fd = ksys_open(name, 0, 0);
-			ksys_ioctl(fd, BLKRRPART, 0);
-		}
 		ksys_close(fd);
 	}
 }
Signature · 1 line
2.26.2
#5 lore raw

PATCH 5/17 md: remove the kernel version of md_u.h

In reply to Christoph Hellwig

mdp_major can just move to drivers/md/md.h.

---
 drivers/md/md.h           |  1 +
 include/linux/raid/md_u.h | 13 -------------
 2 files changed, 1 insertion(+), 13 deletions(-)
 delete mode 100644 include/linux/raid/md_u.h
drivers/md/md.h +1 -0
index 37315a3f28e97d..6f8fff77ce10a5 100644
@@ -801,6 +801,7 @@ static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio
 		mddev->queue->limits.max_write_zeroes_sectors = 0;
 }
 
+extern int mdp_major;
 void md_autostart_arrays(int part);
 
 #endif /* _MD_MD_H */
include/linux/raid/md_u.h deleted +0 -13
deleted file mode 100644
index 8dfec085a20ee1..00000000000000
@@ -1,13 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-or-later */
-/*
-   md_u.h : user <=> kernel API between Linux raidtools and RAID drivers
-          Copyright (C) 1998 Ingo Molnar
-	  
-*/
-#ifndef _MD_U_H
-#define _MD_U_H
-
-#include <uapi/linux/raid/md_u.h>
-
-extern int mdp_major;
-#endif 
Signature · 1 line
2.26.2
#6 lore raw

PATCH 6/17 md: simplify md_setup_drive

In reply to Christoph Hellwig

Move the loop over the possible arrays into the caller to remove a level of indentation for the whole function.

---
 drivers/md/md-autodetect.c | 203 ++++++++++++++++++-------------------
 1 file changed, 101 insertions(+), 102 deletions(-)
drivers/md/md-autodetect.c +101 -102
index 6bc9b734eee6ff..a43a8f1580584c 100644
@@ -27,7 +27,7 @@ static int __initdata raid_noautodetect=1;
 #endif
 static int __initdata raid_autopart;
 
-static struct {
+static struct md_setup_args {
 	int minor;
 	int partitioned;
 	int level;
@@ -126,122 +126,117 @@ static inline int create_dev(char *name, dev_t dev)
 	return ksys_mknod(name, S_IFBLK|0600, new_encode_dev(dev));
 }
 
-static void __init md_setup_drive(void)
+static void __init md_setup_drive(struct md_setup_args *args)
 {
-	int minor, i, ent, partitioned;
+	int minor, i, partitioned;
 	dev_t dev;
 	dev_t devices[MD_SB_DISKS+1];
+	int fd;
+	int err = 0;
+	char *devname;
+	mdu_disk_info_t dinfo;
+	char name[16];
 
-	for (ent = 0; ent < md_setup_ents ; ent++) {
-		int fd;
-		int err = 0;
-		char *devname;
-		mdu_disk_info_t dinfo;
-		char name[16];
+	minor = args->minor;
+	partitioned = args->partitioned;
+	devname = args->device_names;
 
-		minor = md_setup_args[ent].minor;
-		partitioned = md_setup_args[ent].partitioned;
-		devname = md_setup_args[ent].device_names;
+	sprintf(name, "/dev/md%s%d", partitioned?"_d":"", minor);
+	if (partitioned)
+		dev = MKDEV(mdp_major, minor << MdpMinorShift);
+	else
+		dev = MKDEV(MD_MAJOR, minor);
+	create_dev(name, dev);
+	for (i = 0; i < MD_SB_DISKS && devname != NULL; i++) {
+		struct kstat stat;
+		char *p;
+		char comp_name[64];
 
-		sprintf(name, "/dev/md%s%d", partitioned?"_d":"", minor);
-		if (partitioned)
-			dev = MKDEV(mdp_major, minor << MdpMinorShift);
-		else
-			dev = MKDEV(MD_MAJOR, minor);
-		create_dev(name, dev);
-		for (i = 0; i < MD_SB_DISKS && devname != NULL; i++) {
-			struct kstat stat;
-			char *p;
-			char comp_name[64];
+		p = strchr(devname, ',');
+		if (p)
+			*p++ = 0;
 
-			p = strchr(devname, ',');
-			if (p)
-				*p++ = 0;
+		dev = name_to_dev_t(devname);
+		if (strncmp(devname, "/dev/", 5) == 0)
+			devname += 5;
+		snprintf(comp_name, 63, "/dev/%s", devname);
+		if (vfs_stat(comp_name, &stat) == 0 && S_ISBLK(stat.mode))
+			dev = new_decode_dev(stat.rdev);
+		if (!dev) {
+			printk(KERN_WARNING "md: Unknown device name: %s\n", devname);
+			break;
+		}
 
-			dev = name_to_dev_t(devname);
-			if (strncmp(devname, "/dev/", 5) == 0)
-				devname += 5;
-			snprintf(comp_name, 63, "/dev/%s", devname);
-			if (vfs_stat(comp_name, &stat) == 0 &&
-			    S_ISBLK(stat.mode))
-				dev = new_decode_dev(stat.rdev);
-			if (!dev) {
-				printk(KERN_WARNING "md: Unknown device name: %s\n", devname);
-				break;
-			}
+		devices[i] = dev;
+		devname = p;
+	}
+	devices[i] = 0;
 
-			devices[i] = dev;
+	if (!i)
+		return;
 
-			devname = p;
-		}
-		devices[i] = 0;
+	printk(KERN_INFO "md: Loading md%s%d: %s\n",
+		partitioned ? "_d" : "", minor,
+		args->device_names);
 
-		if (!i)
-			continue;
+	fd = ksys_open(name, 0, 0);
+	if (fd < 0) {
+		printk(KERN_ERR "md: open failed - cannot start "
+				"array %s\n", name);
+		return;
+	}
+	if (ksys_ioctl(fd, SET_ARRAY_INFO, 0) == -EBUSY) {
+		printk(KERN_WARNING
+		       "md: Ignoring md=%d, already autodetected. (Use raid=noautodetect)\n",
+		       minor);
+		ksys_close(fd);
+		return;
+	}
 
-		printk(KERN_INFO "md: Loading md%s%d: %s\n",
-			partitioned ? "_d" : "", minor,
-			md_setup_args[ent].device_names);
+	if (args->level != LEVEL_NONE) {
+		/* non-persistent */
+		mdu_array_info_t ainfo;
+		ainfo.level = args->level;
+		ainfo.size = 0;
+		ainfo.nr_disks =0;
+		ainfo.raid_disks =0;
+		while (devices[ainfo.raid_disks])
+			ainfo.raid_disks++;
+		ainfo.md_minor =minor;
+		ainfo.not_persistent = 1;
 
-		fd = ksys_open(name, 0, 0);
-		if (fd < 0) {
-			printk(KERN_ERR "md: open failed - cannot start "
-					"array %s\n", name);
-			continue;
-		}
-		if (ksys_ioctl(fd, SET_ARRAY_INFO, 0) == -EBUSY) {
-			printk(KERN_WARNING
-			       "md: Ignoring md=%d, already autodetected. (Use raid=noautodetect)\n",
-			       minor);
-			ksys_close(fd);
-			continue;
+		ainfo.state = (1 << MD_SB_CLEAN);
+		ainfo.layout = 0;
+		ainfo.chunk_size = args->chunk;
+		err = ksys_ioctl(fd, SET_ARRAY_INFO, (long)&ainfo);
+		for (i = 0; !err && i <= MD_SB_DISKS; i++) {
+			dev = devices[i];
+			if (!dev)
+				break;
+			dinfo.number = i;
+			dinfo.raid_disk = i;
+			dinfo.state = (1<<MD_DISK_ACTIVE)|(1<<MD_DISK_SYNC);
+			dinfo.major = MAJOR(dev);
+			dinfo.minor = MINOR(dev);
+			err = ksys_ioctl(fd, ADD_NEW_DISK,
+					 (long)&dinfo);
 		}
-
-		if (md_setup_args[ent].level != LEVEL_NONE) {
-			/* non-persistent */
-			mdu_array_info_t ainfo;
-			ainfo.level = md_setup_args[ent].level;
-			ainfo.size = 0;
-			ainfo.nr_disks =0;
-			ainfo.raid_disks =0;
-			while (devices[ainfo.raid_disks])
-				ainfo.raid_disks++;
-			ainfo.md_minor =minor;
-			ainfo.not_persistent = 1;
-
-			ainfo.state = (1 << MD_SB_CLEAN);
-			ainfo.layout = 0;
-			ainfo.chunk_size = md_setup_args[ent].chunk;
-			err = ksys_ioctl(fd, SET_ARRAY_INFO, (long)&ainfo);
-			for (i = 0; !err && i <= MD_SB_DISKS; i++) {
-				dev = devices[i];
-				if (!dev)
-					break;
-				dinfo.number = i;
-				dinfo.raid_disk = i;
-				dinfo.state = (1<<MD_DISK_ACTIVE)|(1<<MD_DISK_SYNC);
-				dinfo.major = MAJOR(dev);
-				dinfo.minor = MINOR(dev);
-				err = ksys_ioctl(fd, ADD_NEW_DISK,
-						 (long)&dinfo);
-			}
-		} else {
-			/* persistent */
-			for (i = 0; i <= MD_SB_DISKS; i++) {
-				dev = devices[i];
-				if (!dev)
-					break;
-				dinfo.major = MAJOR(dev);
-				dinfo.minor = MINOR(dev);
-				ksys_ioctl(fd, ADD_NEW_DISK, (long)&dinfo);
-			}
+	} else {
+		/* persistent */
+		for (i = 0; i <= MD_SB_DISKS; i++) {
+			dev = devices[i];
+			if (!dev)
+				break;
+			dinfo.major = MAJOR(dev);
+			dinfo.minor = MINOR(dev);
+			ksys_ioctl(fd, ADD_NEW_DISK, (long)&dinfo);
 		}
-		if (!err)
-			err = ksys_ioctl(fd, RUN_ARRAY, 0);
-		if (err)
-			printk(KERN_WARNING "md: starting md%d failed\n", minor);
-		ksys_close(fd);
 	}
+	if (!err)
+		err = ksys_ioctl(fd, RUN_ARRAY, 0);
+	if (err)
+		printk(KERN_WARNING "md: starting md%d failed\n", minor);
+	ksys_close(fd);
 }
 
 static int __init raid_setup(char *str)
@@ -289,11 +284,15 @@ static void __init autodetect_raid(void)
 
 void __init md_run_setup(void)
 {
+	int ent;
+
 	create_dev("/dev/md0", MKDEV(MD_MAJOR, 0));
 
 	if (raid_noautodetect)
 		printk(KERN_INFO "md: Skipping autodetection of RAID arrays. (raid=autodetect will force)\n");
 	else
 		autodetect_raid();
-	md_setup_drive();
+
+	for (ent = 0; ent < md_setup_ents; ent++)
+		md_setup_drive(&md_setup_args[ent]);
 }
Signature · 1 line
2.26.2
#7 lore raw

PATCH 7/17 md: rewrite md_setup_drive to avoid ioctls

In reply to Christoph Hellwig

md_setup_drive knows it works with md devices, so it is rather pointless to open a file descriptor and issue ioctls. Just call directly into the relevant low-level md routines after getting a handle to the device using blkdev_get_by_dev instead.

---
 drivers/md/md-autodetect.c | 127 ++++++++++++++++---------------------
 drivers/md/md.c            |  20 +++---
 drivers/md/md.h            |   6 ++
 3 files changed, 71 insertions(+), 82 deletions(-)
drivers/md/md-autodetect.c +56 -71
index a43a8f1580584c..5b24b5616d3acc 100644
@@ -2,7 +2,6 @@
 #include <linux/kernel.h>
 #include <linux/blkdev.h>
 #include <linux/init.h>
-#include <linux/syscalls.h>
 #include <linux/mount.h>
 #include <linux/major.h>
 #include <linux/delay.h>
@@ -120,37 +119,29 @@ static int __init md_setup(char *str)
 	return 1;
 }
 
-static inline int create_dev(char *name, dev_t dev)
-{
-	ksys_unlink(name);
-	return ksys_mknod(name, S_IFBLK|0600, new_encode_dev(dev));
-}
-
 static void __init md_setup_drive(struct md_setup_args *args)
 {
-	int minor, i, partitioned;
-	dev_t dev;
-	dev_t devices[MD_SB_DISKS+1];
-	int fd;
-	int err = 0;
-	char *devname;
-	mdu_disk_info_t dinfo;
+	char *devname = args->device_names;
+	dev_t devices[MD_SB_DISKS + 1], mdev;
+	struct mdu_array_info_s ainfo = { };
+	struct block_device *bdev;
+	struct mddev *mddev;
+	int err = 0, i;
 	char name[16];
 
-	minor = args->minor;
-	partitioned = args->partitioned;
-	devname = args->device_names;
+	if (args->partitioned) {
+		mdev = MKDEV(mdp_major, args->minor << MdpMinorShift);
+		sprintf(name, "md_d%d", args->minor);
+	} else {
+		mdev = MKDEV(MD_MAJOR, args->minor);
+		sprintf(name, "md%d", args->minor);
+	}
 
-	sprintf(name, "/dev/md%s%d", partitioned?"_d":"", minor);
-	if (partitioned)
-		dev = MKDEV(mdp_major, minor << MdpMinorShift);
-	else
-		dev = MKDEV(MD_MAJOR, minor);
-	create_dev(name, dev);
 	for (i = 0; i < MD_SB_DISKS && devname != NULL; i++) {
 		struct kstat stat;
 		char *p;
 		char comp_name[64];
+		dev_t dev;
 
 		p = strchr(devname, ',');
 		if (p)
@@ -163,7 +154,7 @@ static void __init md_setup_drive(struct md_setup_args *args)
 		if (vfs_stat(comp_name, &stat) == 0 && S_ISBLK(stat.mode))
 			dev = new_decode_dev(stat.rdev);
 		if (!dev) {
-			printk(KERN_WARNING "md: Unknown device name: %s\n", devname);
+			pr_warn("md: Unknown device name: %s\n", devname);
 			break;
 		}
 
@@ -175,68 +166,64 @@ static void __init md_setup_drive(struct md_setup_args *args)
 	if (!i)
 		return;
 
-	printk(KERN_INFO "md: Loading md%s%d: %s\n",
-		partitioned ? "_d" : "", minor,
-		args->device_names);
+	pr_info("md: Loading %s: %s\n", name, args->device_names);
 
-	fd = ksys_open(name, 0, 0);
-	if (fd < 0) {
-		printk(KERN_ERR "md: open failed - cannot start "
-				"array %s\n", name);
+	bdev = blkdev_get_by_dev(mdev, FMODE_READ, NULL);
+	if (IS_ERR(bdev)) {
+		pr_err("md: open failed - cannot start array %s\n", name);
 		return;
 	}
-	if (ksys_ioctl(fd, SET_ARRAY_INFO, 0) == -EBUSY) {
-		printk(KERN_WARNING
-		       "md: Ignoring md=%d, already autodetected. (Use raid=noautodetect)\n",
-		       minor);
-		ksys_close(fd);
-		return;
+	mddev = bdev->bd_disk->private_data;
+
+	err = mddev_lock(mddev);
+	if (err) {
+		pr_err("md: failed to lock array %s\n", name);
+		goto out_blkdev_put;
+	}
+
+	if (!list_empty(&mddev->disks) || mddev->raid_disks) {
+		pr_warn("md: Ignoring %s, already autodetected. (Use raid=noautodetect)\n",
+		       name);
+		goto out_unlock;
 	}
 
 	if (args->level != LEVEL_NONE) {
 		/* non-persistent */
-		mdu_array_info_t ainfo;
 		ainfo.level = args->level;
-		ainfo.size = 0;
-		ainfo.nr_disks =0;
-		ainfo.raid_disks =0;
-		while (devices[ainfo.raid_disks])
-			ainfo.raid_disks++;
-		ainfo.md_minor =minor;
+		ainfo.md_minor = args->minor;
 		ainfo.not_persistent = 1;
-
 		ainfo.state = (1 << MD_SB_CLEAN);
-		ainfo.layout = 0;
 		ainfo.chunk_size = args->chunk;
-		err = ksys_ioctl(fd, SET_ARRAY_INFO, (long)&ainfo);
-		for (i = 0; !err && i <= MD_SB_DISKS; i++) {
-			dev = devices[i];
-			if (!dev)
-				break;
+		while (devices[ainfo.raid_disks])
+			ainfo.raid_disks++;
+	}
+
+	err = md_set_array_info(mddev, &ainfo);
+
+	for (i = 0; i <= MD_SB_DISKS && devices[i]; i++) {
+		struct mdu_disk_info_s dinfo = {
+			.major	= MAJOR(devices[i]),
+			.minor	= MINOR(devices[i]),
+		};
+
+		if (args->level != LEVEL_NONE) {
 			dinfo.number = i;
 			dinfo.raid_disk = i;
-			dinfo.state = (1<<MD_DISK_ACTIVE)|(1<<MD_DISK_SYNC);
-			dinfo.major = MAJOR(dev);
-			dinfo.minor = MINOR(dev);
-			err = ksys_ioctl(fd, ADD_NEW_DISK,
-					 (long)&dinfo);
-		}
-	} else {
-		/* persistent */
-		for (i = 0; i <= MD_SB_DISKS; i++) {
-			dev = devices[i];
-			if (!dev)
-				break;
-			dinfo.major = MAJOR(dev);
-			dinfo.minor = MINOR(dev);
-			ksys_ioctl(fd, ADD_NEW_DISK, (long)&dinfo);
+			dinfo.state =
+				(1 << MD_DISK_ACTIVE) | (1 << MD_DISK_SYNC);
 		}
+
+		md_add_new_disk(mddev, &dinfo);
 	}
+
 	if (!err)
-		err = ksys_ioctl(fd, RUN_ARRAY, 0);
+		err = do_md_run(mddev);
 	if (err)
-		printk(KERN_WARNING "md: starting md%d failed\n", minor);
-	ksys_close(fd);
+		pr_warn("md: starting %s failed\n", name);
+out_unlock:
+	mddev_unlock(mddev);
+out_blkdev_put:
+	blkdev_put(bdev, FMODE_READ);
 }
 
 static int __init raid_setup(char *str)
@@ -286,8 +273,6 @@ void __init md_run_setup(void)
 {
 	int ent;
 
-	create_dev("/dev/md0", MKDEV(MD_MAJOR, 0));
-
 	if (raid_noautodetect)
 		printk(KERN_INFO "md: Skipping autodetection of RAID arrays. (raid=autodetect will force)\n");
 	else
drivers/md/md.c +9 -11
index 6e9a48da474848..9960cfeb59a50c 100644
@@ -4368,7 +4368,6 @@ array_state_show(struct mddev *mddev, char *page)
 
 static int do_md_stop(struct mddev *mddev, int ro, struct block_device *bdev);
 static int md_set_readonly(struct mddev *mddev, struct block_device *bdev);
-static int do_md_run(struct mddev *mddev);
 static int restart_array(struct mddev *mddev);
 
 static ssize_t
@@ -6015,7 +6014,7 @@ int md_run(struct mddev *mddev)
 }
 EXPORT_SYMBOL_GPL(md_run);
 
-static int do_md_run(struct mddev *mddev)
+int do_md_run(struct mddev *mddev)
 {
 	int err;
 
@@ -6651,7 +6650,7 @@ static int get_disk_info(struct mddev *mddev, void __user * arg)
 	return 0;
 }
 
-static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
+int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info)
 {
 	char b[BDEVNAME_SIZE], b2[BDEVNAME_SIZE];
 	struct md_rdev *rdev;
@@ -6697,7 +6696,7 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 	}
 
 	/*
-	 * add_new_disk can be used once the array is assembled
+	 * md_add_new_disk can be used once the array is assembled
 	 * to add "hot spares".  They must already have a superblock
 	 * written
 	 */
@@ -6810,7 +6809,7 @@ static int add_new_disk(struct mddev *mddev, mdu_disk_info_t *info)
 		return err;
 	}
 
-	/* otherwise, add_new_disk is only allowed
+	/* otherwise, md_add_new_disk is only allowed
 	 * for major_version==0 superblocks
 	 */
 	if (mddev->major_version != 0) {
@@ -7055,7 +7054,7 @@ static int set_bitmap_file(struct mddev *mddev, int fd)
 }
 
 /*
- * set_array_info is used two different ways
+ * md_set_array_info is used two different ways
  * The original usage is when creating a new array.
  * In this usage, raid_disks is > 0 and it together with
  *  level, size, not_persistent,layout,chunksize determine the
@@ -7067,9 +7066,8 @@ static int set_bitmap_file(struct mddev *mddev, int fd)
  *  The minor and patch _version numbers are also kept incase the
  *  super_block handler wishes to interpret them.
  */
-static int set_array_info(struct mddev *mddev, mdu_array_info_t *info)
+int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info)
 {
-
 	if (info->raid_disks == 0) {
 		/* just setting version number for superblock loading */
 		if (info->major_version < 0 ||
@@ -7560,7 +7558,7 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 			err = -EBUSY;
 			goto unlock;
 		}
-		err = set_array_info(mddev, &info);
+		err = md_set_array_info(mddev, &info);
 		if (err) {
 			pr_warn("md: couldn't set array info. %d\n", err);
 			goto unlock;
@@ -7614,7 +7612,7 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 				/* Need to clear read-only for this */
 				break;
 			else
-				err = add_new_disk(mddev, &info);
+				err = md_add_new_disk(mddev, &info);
 			goto unlock;
 		}
 		break;
@@ -7682,7 +7680,7 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 		if (copy_from_user(&info, argp, sizeof(info)))
 			err = -EFAULT;
 		else
-			err = add_new_disk(mddev, &info);
+			err = md_add_new_disk(mddev, &info);
 		goto unlock;
 	}
 
drivers/md/md.h +6 -0
index 6f8fff77ce10a5..7ee81aa2eac862 100644
@@ -801,7 +801,13 @@ static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio
 		mddev->queue->limits.max_write_zeroes_sectors = 0;
 }
 
+struct mdu_array_info_s;
+struct mdu_disk_info_s;
+
 extern int mdp_major;
 void md_autostart_arrays(int part);
+int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info);
+int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info);
+int do_md_run(struct mddev *mddev);
 
 #endif /* _MD_MD_H */
Signature · 1 line
2.26.2
#8 lore raw

PATCH 8/17 initrd: remove support for multiple floppies

In reply to Christoph Hellwig

Remove the special handling for multiple floppies in the initrd code. No one should be using floppies for booting these days. (famous last words..)

---
 arch/arm/kernel/atags_parse.c |  2 -
 arch/sh/kernel/setup.c        |  2 -
 arch/sparc/kernel/setup_32.c  |  2 -
 arch/sparc/kernel/setup_64.c  |  2 -
 arch/x86/kernel/setup.c       |  2 -
 include/linux/initrd.h        |  6 ---
 init/do_mounts.c              | 69 ++++-------------------------------
 init/do_mounts.h              |  1 -
 init/do_mounts_rd.c           | 20 +++-------
 9 files changed, 12 insertions(+), 94 deletions(-)
arch/arm/kernel/atags_parse.c +0 -2
index ce02f92f4ab262..6c12d9fe694e3e 100644
@@ -91,8 +91,6 @@ __tagtable(ATAG_VIDEOTEXT, parse_tag_videotext);
 static int __init parse_tag_ramdisk(const struct tag *tag)
 {
 	rd_image_start = tag->u.ramdisk.start;
-	rd_doload = (tag->u.ramdisk.flags & 1) == 0;
-	rd_prompt = (tag->u.ramdisk.flags & 2) == 0;
 
 	if (tag->u.ramdisk.size)
 		rd_size = tag->u.ramdisk.size;
arch/sh/kernel/setup.c +0 -2
index 67f5a3b44c2eff..4144be650d4106 100644
@@ -290,8 +290,6 @@ void __init setup_arch(char **cmdline_p)
 
 #ifdef CONFIG_BLK_DEV_RAM
 	rd_image_start = RAMDISK_FLAGS & RAMDISK_IMAGE_START_MASK;
-	rd_prompt = ((RAMDISK_FLAGS & RAMDISK_PROMPT_FLAG) != 0);
-	rd_doload = ((RAMDISK_FLAGS & RAMDISK_LOAD_FLAG) != 0);
 #endif
 
 	if (!MOUNT_ROOT_RDONLY)
arch/sparc/kernel/setup_32.c +0 -2
index 6d07b85b9e2470..eea43a1aef1b9a 100644
@@ -353,8 +353,6 @@ void __init setup_arch(char **cmdline_p)
 	ROOT_DEV = old_decode_dev(root_dev);
 #ifdef CONFIG_BLK_DEV_RAM
 	rd_image_start = ram_flags & RAMDISK_IMAGE_START_MASK;
-	rd_prompt = ((ram_flags & RAMDISK_PROMPT_FLAG) != 0);
-	rd_doload = ((ram_flags & RAMDISK_LOAD_FLAG) != 0);	
 #endif
 
 	prom_setsync(prom_sync_me);
arch/sparc/kernel/setup_64.c +0 -2
index f765fda871eb61..d87244197d5cbb 100644
@@ -659,8 +659,6 @@ void __init setup_arch(char **cmdline_p)
 	ROOT_DEV = old_decode_dev(root_dev);
 #ifdef CONFIG_BLK_DEV_RAM
 	rd_image_start = ram_flags & RAMDISK_IMAGE_START_MASK;
-	rd_prompt = ((ram_flags & RAMDISK_PROMPT_FLAG) != 0);
-	rd_doload = ((ram_flags & RAMDISK_LOAD_FLAG) != 0);
 #endif
 
 	task_thread_info(&init_task)->kregs = &fake_swapper_regs;
arch/x86/kernel/setup.c +0 -2
index a3767e74c758c0..b9a68d8e06d8d1 100644
@@ -870,8 +870,6 @@ void __init setup_arch(char **cmdline_p)
 
 #ifdef CONFIG_BLK_DEV_RAM
 	rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
-	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
-	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 #endif
 #ifdef CONFIG_EFI
 	if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
include/linux/initrd.h +0 -6
index aa591435572868..8db6f8c8030b68 100644
@@ -2,12 +2,6 @@
 
 #define INITRD_MINOR 250 /* shouldn't collide with /dev/ram* too soon ... */
 
-/* 1 = load ramdisk, 0 = don't load */
-extern int rd_doload;
-
-/* 1 = prompt for ramdisk, 0 = don't prompt */
-extern int rd_prompt;
-
 /* starting block # of image */
 extern int rd_image_start;
 
init/do_mounts.c +7 -62

107 lines not on this page · raw message

init/do_mounts.h +0 -1
index 50d6c8941e15a1..c855b3f0e06d19 100644
@@ -9,7 +9,6 @@
 #include <linux/major.h>
 #include <linux/root_dev.h>
 
-void  change_floppy(char *fmt, ...);
 void  mount_block_root(char *name, int flags);
 void  mount_root(void);
 extern int root_mountflags;
init/do_mounts_rd.c +5 -15
index 32fb049d18f9b4..27b1bccf6f12a8 100644
@@ -15,11 +15,9 @@
 #include <linux/decompress/generic.h>
 
 
-int __initdata rd_prompt = 1;/* 1 = prompt for RAM disk, 0 = don't prompt */
-
 static int __init prompt_ramdisk(char *str)
 {
-	rd_prompt = simple_strtol(str,NULL,0) & 1;
+	pr_warn("ignoring the depreated prompt_ramdisk= option\n");
 	return 1;
 }
 __setup("prompt_ramdisk=", prompt_ramdisk);
@@ -178,7 +176,7 @@ int __init rd_load_image(char *from)
 	int res = 0;
 	int in_fd, out_fd;
 	unsigned long rd_blocks, devblocks;
-	int nblocks, i, disk;
+	int nblocks, i;
 	char *buf = NULL;
 	unsigned short rotate = 0;
 	decompress_fn decompressor = NULL;
@@ -243,21 +241,15 @@ int __init rd_load_image(char *from)
 
 	printk(KERN_NOTICE "RAMDISK: Loading %dKiB [%ld disk%s] into ram disk... ",
 		nblocks, ((nblocks-1)/devblocks)+1, nblocks>devblocks ? "s" : "");
-	for (i = 0, disk = 1; i < nblocks; i++) {
+	for (i = 0; i < nblocks; i++) {
 		if (i && (i % devblocks == 0)) {
-			pr_cont("done disk #%d.\n", disk++);
+			pr_cont("done disk #1.\n");
 			rotate = 0;
 			if (ksys_close(in_fd)) {
 				printk("Error closing the disk.\n");
 				goto noclose_input;
 			}
-			change_floppy("disk #%d", disk);
-			in_fd = ksys_open(from, O_RDONLY, 0);
-			if (in_fd < 0)  {
-				printk("Error opening disk.\n");
-				goto noclose_input;
-			}
-			printk("Loading disk #%d... ", disk);
+			break;
 		}
 		ksys_read(in_fd, buf, BLOCK_SIZE);
 		ksys_write(out_fd, buf, BLOCK_SIZE);
@@ -284,8 +276,6 @@ int __init rd_load_image(char *from)
 
 int __init rd_load_disk(int n)
 {
-	if (rd_prompt)
-		change_floppy("root floppy disk to be loaded into RAM disk");
 	create_dev("/dev/root", ROOT_DEV);
 	create_dev("/dev/ram", MKDEV(RAMDISK_MAJOR, n));
 	return rd_load_image("/dev/root");
Signature · 1 line
2.26.2
#9 lore raw

PATCH 9/17 initrd: remove the BLKFLSBUF call in handle_initrd

In reply to Christoph Hellwig

BLKFLSBUF used to be overloaded for the ramdisk driver to free the whole ramdisk, which was completely different behavior compared to all other drivers. But this magic overload got removed in commit ff26956875c2 ("brd: remove support for BLKFLSBUF"), so this call is entirely pointless now.

---
 init/do_mounts_initrd.c | 9 ---------
 1 file changed, 9 deletions(-)
init/do_mounts_initrd.c +0 -9
index d72beda824aa79..e4f88e9e1c0839 100644
@@ -115,21 +115,12 @@ static void __init handle_initrd(void)
 	if (!error)
 		printk("okay\n");
 	else {
-		int fd = ksys_open("/dev/root.old", O_RDWR, 0);
 		if (error == -ENOENT)
 			printk("/initrd does not exist. Ignored.\n");
 		else
 			printk("failed\n");
 		printk(KERN_NOTICE "Unmounting old root\n");
 		ksys_umount("/old", MNT_DETACH);
-		printk(KERN_NOTICE "Trying to free ramdisk memory ... ");
-		if (fd < 0) {
-			error = fd;
-		} else {
-			error = ksys_ioctl(fd, BLKFLSBUF, 0);
-			ksys_close(fd);
-		}
-		printk(!error ? "okay\n" : "failed\n");
 	}
 }
 
Signature · 1 line
2.26.2
#10 lore raw

PATCH 10/17 initrd: switch initrd loading to struct file based APIs

In reply to Christoph Hellwig

There is no good reason to mess with file descriptors from in-kernel code, switch the initrd loading to struct file based read and writes instead.

---
 fs/read_write.c          |  2 +-
 include/linux/syscalls.h |  1 -
 init/do_mounts_rd.c      | 82 ++++++++++++++++++++--------------------
 3 files changed, 43 insertions(+), 42 deletions(-)
fs/read_write.c +1 -1
index bbfa9b12b15eb7..fc1cbe2517894f 100644
@@ -301,7 +301,7 @@ loff_t vfs_llseek(struct file *file, loff_t offset, int whence)
 }
 EXPORT_SYMBOL(vfs_llseek);
 
-off_t ksys_lseek(unsigned int fd, off_t offset, unsigned int whence)
+static off_t ksys_lseek(unsigned int fd, off_t offset, unsigned int whence)
 {
 	off_t retval;
 	struct fd f = fdget_pos(fd);
include/linux/syscalls.h +0 -1
index b951a87da9877c..10843a6adb770d 100644
@@ -1246,7 +1246,6 @@ int ksys_fchown(unsigned int fd, uid_t user, gid_t group);
 int ksys_getdents64(unsigned int fd, struct linux_dirent64 __user *dirent,
 		    unsigned int count);
 int ksys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
-off_t ksys_lseek(unsigned int fd, off_t offset, unsigned int whence);
 ssize_t ksys_read(unsigned int fd, char __user *buf, size_t count);
 void ksys_sync(void);
 int ksys_unshare(unsigned long unshare_flags);
init/do_mounts_rd.c +42 -40

197 lines not on this page · raw message

Signature · 1 line
2.26.2
#11 lore raw

PATCH 11/17 initrd: mark init_linuxrc as __init

In reply to Christoph Hellwig
---
 init/do_mounts_initrd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
init/do_mounts_initrd.c +1 -1
index e4f88e9e1c0839..57ad5b2da8f5f5 100644
@@ -56,7 +56,7 @@ static int __init early_initrd(char *p)
 }
 early_param("initrd", early_initrd);
 
-static int init_linuxrc(struct subprocess_info *info, struct cred *new)
+static int __init init_linuxrc(struct subprocess_info *info, struct cred *new)
 {
 	ksys_unshare(CLONE_FS | CLONE_FILES);
 	console_on_rootfs();
Signature · 1 line
2.26.2
#12 lore raw

PATCH 12/17 initrd: mark initrd support as deprecated

In reply to Christoph Hellwig

The classic initial ramdisk has been replaced by the much more flexible and efficient initramfs a long time. Warn about it being removed soon.

---
 init/do_mounts_initrd.c | 2 ++
 1 file changed, 2 insertions(+)
init/do_mounts_initrd.c +2 -0
index 57ad5b2da8f5f5..e08669187d63be 100644
@@ -75,6 +75,8 @@ static void __init handle_initrd(void)
 	extern char *envp_init[];
 	int error;
 
+	pr_warn("using deprecated initrd support, will be removed in 2021.\n");
+
 	real_root_dev = new_encode_dev(ROOT_DEV);
 	create_dev("/dev/root.old", Root_RAM0);
 	/* mount initrd on rootfs' /root */
Signature · 1 line
2.26.2
#13 lore raw

PATCH 13/17 initramfs: remove the populate_initrd_image and clean_rootfs stubs

In reply to Christoph Hellwig

If initrd support is not enable just print the warning directly instead of hiding the fact that we just failed behind two stub functions.

---
 init/initramfs.c | 15 ++++-----------
 1 file changed, 4 insertions(+), 11 deletions(-)
init/initramfs.c +4 -11
index 7a38012e1af742..d10404625c31f0 100644
@@ -618,13 +618,7 @@ static void __init clean_rootfs(void)
 	ksys_close(fd);
 	kfree(buf);
 }
-#else
-static inline void clean_rootfs(void)
-{
-}
-#endif /* CONFIG_BLK_DEV_RAM */
 
-#ifdef CONFIG_BLK_DEV_RAM
 static void __init populate_initrd_image(char *err)
 {
 	ssize_t written;
@@ -644,11 +638,6 @@ static void __init populate_initrd_image(char *err)
 		       written, initrd_end - initrd_start);
 	ksys_close(fd);
 }
-#else
-static void __init populate_initrd_image(char *err)
-{
-	printk(KERN_EMERG "Initramfs unpacking failed: %s\n", err);
-}
 #endif /* CONFIG_BLK_DEV_RAM */
 
 static int __init populate_rootfs(void)
@@ -668,8 +657,12 @@ static int __init populate_rootfs(void)
 
 	err = unpack_to_rootfs((char *)initrd_start, initrd_end - initrd_start);
 	if (err) {
+#ifdef CONFIG_BLK_DEV_RAM
 		clean_rootfs();
 		populate_initrd_image(err);
+#else
+		printk(KERN_EMERG "Initramfs unpacking failed: %s\n", err);
+#endif
 	}
 
 done:
Signature · 1 line
2.26.2
#14 lore raw

PATCH 14/17 initramfs: simplify clean_rootfs

In reply to Christoph Hellwig

Just use d_genocide instead of iterating through the root directory with cumbersome userspace-like APIs. This also ensures we actually remove files that are not direct children of the root entry, which the old code failed to do.

Fixes df52092f3c97 ("fastboot: remove duplicate unpack_to_rootfs()")
---
 fs/readdir.c             | 11 ++--------
 include/linux/syscalls.h |  2 --
 init/initramfs.c         | 46 +++++-----------------------------------
 3 files changed, 7 insertions(+), 52 deletions(-)
fs/readdir.c +2 -9
index a49f07c11cfbd0..19434b3c982cd3 100644
@@ -348,8 +348,8 @@ static int filldir64(struct dir_context *ctx, const char *name, int namlen,
 	return -EFAULT;
 }
 
-int ksys_getdents64(unsigned int fd, struct linux_dirent64 __user *dirent,
-		    unsigned int count)
+SYSCALL_DEFINE3(getdents64, unsigned int, fd,
+		struct linux_dirent64 __user *, dirent, unsigned int, count)
 {
 	struct fd f;
 	struct getdents_callback64 buf = {
@@ -380,13 +380,6 @@ int ksys_getdents64(unsigned int fd, struct linux_dirent64 __user *dirent,
 	return error;
 }
 
-
-SYSCALL_DEFINE3(getdents64, unsigned int, fd,
-		struct linux_dirent64 __user *, dirent, unsigned int, count)
-{
-	return ksys_getdents64(fd, dirent, count);
-}
-
 #ifdef CONFIG_COMPAT
 struct compat_old_linux_dirent {
 	compat_ulong_t	d_ino;
include/linux/syscalls.h +0 -2
index 10843a6adb770d..a998651629c71b 100644
@@ -1243,8 +1243,6 @@ ssize_t ksys_write(unsigned int fd, const char __user *buf, size_t count);
 int ksys_chdir(const char __user *filename);
 int ksys_fchmod(unsigned int fd, umode_t mode);
 int ksys_fchown(unsigned int fd, uid_t user, gid_t group);
-int ksys_getdents64(unsigned int fd, struct linux_dirent64 __user *dirent,
-		    unsigned int count);
 int ksys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
 ssize_t ksys_read(unsigned int fd, char __user *buf, size_t count);
 void ksys_sync(void);
init/initramfs.c +5 -41

62 lines not on this page · raw message

Signature · 1 line
2.26.2
#15 lore raw

PATCH 15/17 initramfs: switch initramfs unpacking to struct file based APIs

In reply to Christoph Hellwig

There is no good reason to mess with file descriptors from in-kernel code, switch the initramfs unpacking to struct file based write instead. As we don't have nice helper for chmod or chown on a struct file or struct path use the pathname based ones instead there. This causes additional (cached) lookups, but keeps the code much simpler.

---
 fs/open.c                |  7 +------
 include/linux/syscalls.h |  1 -
 init/initramfs.c         | 42 ++++++++++++++++++++--------------------
 3 files changed, 22 insertions(+), 28 deletions(-)
fs/open.c +1 -6
index 6cd48a61cda3b9..6348173532e663 100644
@@ -602,7 +602,7 @@ static int chmod_common(const struct path *path, umode_t mode)
 	return error;
 }
 
-int ksys_fchmod(unsigned int fd, umode_t mode)
+SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
 {
 	struct fd f = fdget(fd);
 	int err = -EBADF;
@@ -615,11 +615,6 @@ int ksys_fchmod(unsigned int fd, umode_t mode)
 	return err;
 }
 
-SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
-{
-	return ksys_fchmod(fd, mode);
-}
-
 int do_fchmodat(int dfd, const char __user *filename, umode_t mode)
 {
 	struct path path;
include/linux/syscalls.h +0 -1
index a998651629c71b..e8810d75efa9b0 100644
@@ -1241,7 +1241,6 @@ int ksys_dup(unsigned int fildes);
 int ksys_chroot(const char __user *filename);
 ssize_t ksys_write(unsigned int fd, const char __user *buf, size_t count);
 int ksys_chdir(const char __user *filename);
-int ksys_fchmod(unsigned int fd, umode_t mode);
 int ksys_fchown(unsigned int fd, uid_t user, gid_t group);
 int ksys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
 ssize_t ksys_read(unsigned int fd, char __user *buf, size_t count);
init/initramfs.c +21 -21

94 lines not on this page · raw message

Signature · 1 line
2.26.2
#16 lore raw

Re: [PATCH 15/17] initramfs: switch initramfs unpacking to struct file based APIs

Quoted 6 lines · Christoph Hellwig

There is no good reason to mess with file descriptors from in-kernel code, switch the initramfs unpacking to struct file based write instead. As we don't have nice helper for chmod or chown on a struct file or struct path use the pathname based ones instead there. This causes additional (cached) lookups, but keeps the code much simpler.

This is the only one I'm not a huge fan of.

I agree about moving to 'struct file'. But then you could just do the chown/chmod using chown/chmod_common() on file->f_path.

That would keep the same semantics, and it feels like a more straightforward patch.

It would still remove the nasty ksys_fchmod/fchmod, it would just require our - already existing - *_common() functions to be non-static (and maybe renamed to "vfs_chown/chmod()" instead, that "*_common()" naming looks a bit odd compared to all our other "vfs_operation()" helpers).

Linus

#17 lore raw

Re: [PATCH 15/17] initramfs: switch initramfs unpacking to struct file based APIs

Quoted 13 lines · Linus Torvalds
Quoted 6 lines · Christoph Hellwig

There is no good reason to mess with file descriptors from in-kernel code, switch the initramfs unpacking to struct file based write instead. As we don't have nice helper for chmod or chown on a struct file or struct path use the pathname based ones instead there. This causes additional (cached) lookups, but keeps the code much simpler.

This is the only one I'm not a huge fan of.

I agree about moving to 'struct file'. But then you could just do the chown/chmod using chown/chmod_common() on file->f_path.

Quoted 8 lines

That would keep the same semantics, and it feels like a more straightforward patch.

It would still remove the nasty ksys_fchmod/fchmod, it would just require our - already existing - *_common() functions to be non-static (and maybe renamed to "vfs_chown/chmod()" instead, that "*_common()" naming looks a bit odd compared to all our other "vfs_operation()" helpers).

Sure, we can do that. It requires a little more boilerplate that I thought we could just skip.

#18 lore raw

Re: [PATCH 15/17] initramfs: switch initramfs unpacking to struct file based APIs

Quoted 18 lines · Christoph Hellwig
Quoted 16 lines · Linus Torvalds
Quoted 6 lines · Christoph Hellwig

There is no good reason to mess with file descriptors from in-kernel code, switch the initramfs unpacking to struct file based write instead. As we don't have nice helper for chmod or chown on a struct file or struct path use the pathname based ones instead there. This causes additional (cached) lookups, but keeps the code much simpler.

This is the only one I'm not a huge fan of.

I agree about moving to 'struct file'. But then you could just do the chown/chmod using chown/chmod_common() on file->f_path.

That would keep the same semantics, and it feels like a more straightforward patch.

Quoted 8 lines
Quoted 5 lines

It would still remove the nasty ksys_fchmod/fchmod, it would just require our - already existing - *_common() functions to be non-static (and maybe renamed to "vfs_chown/chmod()" instead, that "*_common()" naming looks a bit odd compared to all our other "vfs_operation()" helpers).

Sure, we can do that. It requires a little more boilerplate that I thought we could just skip.

Yeah, it makes the patch perhaps a bit larger (or rather, it adds one new patch to do that "introduce new names for *_common() and add them to the headers"), but just looking at your current patch:

+                       ksys_chown(collected, uid, gid);
+                       ksys_chmod(collected, mode);
+                       if (body_len)

+ vfs_truncate(&wfile->f_path, body_len);

and I go "wouldn't that look a _lot_ more regular something like this":

+                       vfs_chown(&wfile->f_path, uid, gid);
+                       vfs_chmod(&wfile->f_path, mode);
+                       if (body_len)

+ vfs_truncate(&wfile->f_path, body_len);

so I think that I'd prefer it simply because it makes all our interfaces much more natural.

I note that we still need that ksys_chown() because of disgusting issues with the legacy 16-bit uid support. Which is likely not used any more now that a.out support is gone (well, alpha and m68k still define HAVE_AOUT, so it's not "gone" gone, but it's gone).

I can't imagine that anybody would have ELF and 16-bit uid's.

Except users are crazy. So who knows. But it might definitely be time to deprecate UID16.

That's a separate issue, though, it just came up because of that ksys_chown().

Linus

#19 lore raw

PATCH 16/17 init: open code setting up stdin/stdout/stderr

In reply to Christoph Hellwig

Don't rely on the implicit set_fs(KERNEL_DS) for ksys_open to work, but instead open a struct file for /dev/console and then install it as FD 0/1/2 manually.

---
 fs/file.c                |  7 +------
 include/linux/syscalls.h |  1 -
 init/main.c              | 16 ++++++++++------
 3 files changed, 11 insertions(+), 13 deletions(-)
fs/file.c +1 -6
index abb8b7081d7a44..85b7993165dd2f 100644
@@ -985,7 +985,7 @@ SYSCALL_DEFINE2(dup2, unsigned int, oldfd, unsigned int, newfd)
 	return ksys_dup3(oldfd, newfd, 0);
 }
 
-int ksys_dup(unsigned int fildes)
+SYSCALL_DEFINE1(dup, unsigned int, fildes)
 {
 	int ret = -EBADF;
 	struct file *file = fget_raw(fildes);
@@ -1000,11 +1000,6 @@ int ksys_dup(unsigned int fildes)
 	return ret;
 }
 
-SYSCALL_DEFINE1(dup, unsigned int, fildes)
-{
-	return ksys_dup(fildes);
-}
-
 int f_dupfd(unsigned int from, struct file *file, unsigned flags)
 {
 	int err;
include/linux/syscalls.h +0 -1
index e8810d75efa9b0..59af62090ac400 100644
@@ -1237,7 +1237,6 @@ asmlinkage long sys_ni_syscall(void);
  */
 
 int ksys_umount(char __user *name, int flags);
-int ksys_dup(unsigned int fildes);
 int ksys_chroot(const char __user *filename);
 ssize_t ksys_write(unsigned int fd, const char __user *buf, size_t count);
 int ksys_chdir(const char __user *filename);
init/main.c +10 -6
index 0ead83e86b5aa2..db0621dfbb0468 100644
@@ -1457,15 +1457,19 @@ static int __ref kernel_init(void *unused)
 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
 }
 
+/* Open /dev/console, for stdin/stdout/stderr, this should never fail */
 void console_on_rootfs(void)
 {
-	/* Open the /dev/console as stdin, this should never fail */
-	if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
-		pr_err("Warning: unable to open an initial console.\n");
+	struct file *file = filp_open("/dev/console", O_RDWR, 0);
 
-	/* create stdout/stderr */
-	(void) ksys_dup(0);
-	(void) ksys_dup(0);
+	if (IS_ERR(file)) {
+		pr_err("Warning: unable to open an initial console.\n");
+		return;
+	}
+	get_file_rcu_many(file, 2);
+	fd_install(get_unused_fd_flags(0), file);
+	fd_install(get_unused_fd_flags(0), file);
+	fd_install(get_unused_fd_flags(0), file);
 }
 
 static noinline void __init kernel_init_freeable(void)
Signature · 1 line
2.26.2
#20 lore raw

Re: [PATCH 16/17] init: open code setting up stdin/stdout/stderr

Quoted 26 lines · Christoph Hellwig

Don't rely on the implicit set_fs(KERNEL_DS) for ksys_open to work, but instead open a struct file for /dev/console and then install it as FD 0/1/2 manually.

---
 fs/file.c                |  7 +------
 include/linux/syscalls.h |  1 -
 init/main.c              | 16 ++++++++++------
 3 files changed, 11 insertions(+), 13 deletions(-)
fs/file.c +1 -1
index abb8b7081d7a44..85b7993165dd2f 100644
@@ -985,7 +985,7 @@ SYSCALL_DEFINE2(dup2, unsigned int, oldfd, unsigned int, newfd)
        return ksys_dup3(oldfd, newfd, 0);
 }
 
-int ksys_dup(unsigned int fildes)
+SYSCALL_DEFINE1(dup, unsigned int, fildes)
 {
        int ret = -EBADF;
        struct file *file = fget_raw(fildes);

fs/file.c int ksys_dup(unsigned int fildes) · line 1007

Quoted 8 lines
fs/file.c +0 -5
@@ -1000,11 +1000,6 @@ int ksys_dup(unsigned int fildes)
        return ret;
 }
 
-SYSCALL_DEFINE1(dup, unsigned int, fildes)
-{
-       return ksys_dup(fildes);
-}
-

Please split the removal of the now unused ksys_*() functions into a separate patch.

--
Brian Gerst

#21 lore raw

PATCH 17/17 fs: remove ksys_open

In reply to Christoph Hellwig

Just open code it in the two callers.

---
 fs/open.c                | 11 ++++++++---
 include/linux/syscalls.h | 11 -----------
 2 files changed, 8 insertions(+), 14 deletions(-)
fs/open.c +8 -3
index 6348173532e663..4375a5a8e726ea 100644
@@ -1193,7 +1193,9 @@ long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
 
 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
 {
-	return ksys_open(filename, flags, mode);
+	if (force_o_largefile())
+		flags |= O_LARGEFILE;
+	return do_sys_open(AT_FDCWD, filename, flags, mode);
 }
 
 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
@@ -1255,9 +1257,12 @@ COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, fla
  */
 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
 {
-	return ksys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
-}
+	int flags = O_CREAT | O_WRONLY | O_TRUNC;
 
+	if (force_o_largefile())
+		flags |= O_LARGEFILE;
+	return do_sys_open(AT_FDCWD, pathname, flags, mode);
+}
 #endif
 
 /*
include/linux/syscalls.h +0 -11
index 59af62090ac400..39ff738997a172 100644
@@ -1372,17 +1372,6 @@ static inline int ksys_close(unsigned int fd)
 	return __close_fd(current->files, fd);
 }
 
-extern long do_sys_open(int dfd, const char __user *filename, int flags,
-			umode_t mode);
-
-static inline long ksys_open(const char __user *filename, int flags,
-			     umode_t mode)
-{
-	if (force_o_largefile())
-		flags |= O_LARGEFILE;
-	return do_sys_open(AT_FDCWD, filename, flags, mode);
-}
-
 extern long do_sys_truncate(const char __user *pathname, loff_t length);
 
 static inline long ksys_truncate(const char __user *pathname, loff_t length)
Signature · 1 line
2.26.2
#22 lore raw
In reply to Christoph Hellwig
Quoted 3 lines · Christoph Hellwig

There is no really good tree for this, so if there are no objections I'd like to set up a new one for linux-next.

All looks good to me. I had a wish-list change for one of the patches that I sent a reply out for, but even without that it's clearly an improvement.

Of course, I just looked at the patches for sanity, rather than testing anything. Maybe there's something stupid in there. But it all looked straightforward. So Ack from me, with the hope that you'd do that "vfs_chown/chmod()" thing.

Linus

#23 lore raw
In reply to Christoph Hellwig
Quoted 42 lines · Christoph Hellwig

Hi all,

this series starts to move the early init code away from requiring KERNEL_DS to be implicitly set during early startup. It does so by first removing legacy unused cruft, and the switches away the code from struct file based APIs to our more usual in-kernel APIs.

There is no really good tree for this, so if there are no objections I'd like to set up a new one for linux-next.

Git tree:

git://git.infradead.org/users/hch/misc.git init-user-pointers

Gitweb:

http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/init-user-pointers

Changes since v1:
- add a patch to deprecated "classic" initrd support

Diffstat:

b/arch/arm/kernel/atags_parse.c |    2 
b/arch/sh/kernel/setup.c        |    2 
b/arch/sparc/kernel/setup_32.c  |    2 
b/arch/sparc/kernel/setup_64.c  |    2 
b/arch/x86/kernel/setup.c       |    2 
b/drivers/md/Makefile           |    3 
b/drivers/md/md-autodetect.c    |  239
++++++++++++++++++----------------------
b/drivers/md/md.c               |   34 +----
b/drivers/md/md.h               |   10 +
b/fs/file.c                     |    7 -
b/fs/open.c                     |   18 +--
b/fs/read_write.c               |    2 
b/fs/readdir.c                  |   11 -
b/include/linux/initrd.h        |    6 -
b/include/linux/raid/detect.h   |    8 +
b/include/linux/syscalls.h      |   16 --
b/init/Makefile                 |    1 
Quoted 8 lines
b/init/do_mounts.c              |   70 +----------
b/init/do_mounts.h              |   21 ---
b/init/do_mounts_initrd.c       |   13 --
b/init/do_mounts_rd.c           |  102 +++++++----------
b/init/initramfs.c              |  103 +++++------------
b/init/main.c                   |   16 +-
include/linux/raid/md_u.h       |   13 --
24 files changed, 251 insertions(+), 452 deletions(-)

I guess I could say something here... ;)

Signature · 1 line
Sent from my Android device with K-9 Mail. Please excuse my brevity.
#24 lore raw
Quoted 52 lines · On Thu, Jul 09, 2020 at 04:32:07PM -0700, [email protected] wrote:
Quoted 42 lines · Christoph Hellwig

Hi all,

this series starts to move the early init code away from requiring KERNEL_DS to be implicitly set during early startup. It does so by first removing legacy unused cruft, and the switches away the code from struct file based APIs to our more usual in-kernel APIs.

There is no really good tree for this, so if there are no objections I'd like to set up a new one for linux-next.

Git tree:

git://git.infradead.org/users/hch/misc.git init-user-pointers

Gitweb:

http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/init-user-pointers

Changes since v1:
- add a patch to deprecated "classic" initrd support

Diffstat:

b/arch/arm/kernel/atags_parse.c |    2 
b/arch/sh/kernel/setup.c        |    2 
b/arch/sparc/kernel/setup_32.c  |    2 
b/arch/sparc/kernel/setup_64.c  |    2 
b/arch/x86/kernel/setup.c       |    2 
b/drivers/md/Makefile           |    3 
b/drivers/md/md-autodetect.c    |  239
++++++++++++++++++----------------------
b/drivers/md/md.c               |   34 +----
b/drivers/md/md.h               |   10 +
b/fs/file.c                     |    7 -
b/fs/open.c                     |   18 +--
b/fs/read_write.c               |    2 
b/fs/readdir.c                  |   11 -
b/include/linux/initrd.h        |    6 -
b/include/linux/raid/detect.h   |    8 +
b/include/linux/syscalls.h      |   16 --
b/init/Makefile                 |    1 
Quoted 8 lines
b/init/do_mounts.c              |   70 +----------
b/init/do_mounts.h              |   21 ---
b/init/do_mounts_initrd.c       |   13 --
b/init/do_mounts_rd.c           |  102 +++++++----------
b/init/initramfs.c              |  103 +++++------------
b/init/main.c                   |   16 +-
include/linux/raid/md_u.h       |   13 --
24 files changed, 251 insertions(+), 452 deletions(-)
Quoted 1 line

I guess I could say something here... ;)

Like adding an ACK? :)

maintainer Acked-by
#25 lore raw
Quoted 54 lines · Christoph Hellwig
Quoted 52 lines · On Thu, Jul 09, 2020 at 04:32:07PM -0700, [email protected] wrote:
Quoted 50 lines · Christoph Hellwig

Hi all,

this series starts to move the early init code away from requiring KERNEL_DS to be implicitly set during early startup. It does so by first removing legacy unused cruft, and the switches away the code from struct file based APIs to our more usual in-kernel APIs.

There is no really good tree for this, so if there are no objections I'd like to set up a new one for linux-next.

Git tree:

git://git.infradead.org/users/hch/misc.git init-user-pointers

Gitweb:

http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/init-user-pointers

Changes since v1:
- add a patch to deprecated "classic" initrd support

Diffstat:

b/arch/arm/kernel/atags_parse.c |    2
b/arch/sh/kernel/setup.c        |    2
b/arch/sparc/kernel/setup_32.c  |    2
b/arch/sparc/kernel/setup_64.c  |    2
b/arch/x86/kernel/setup.c       |    2
b/drivers/md/Makefile           |    3
b/drivers/md/md-autodetect.c    |  239
++++++++++++++++++----------------------
b/drivers/md/md.c               |   34 +----
b/drivers/md/md.h               |   10 +
b/fs/file.c                     |    7 -
b/fs/open.c                     |   18 +--
b/fs/read_write.c               |    2
b/fs/readdir.c                  |   11 -
b/include/linux/initrd.h        |    6 -
b/include/linux/raid/detect.h   |    8 +
b/include/linux/syscalls.h      |   16 --
b/init/Makefile                 |    1
b/init/do_mounts.c              |   70 +----------
b/init/do_mounts.h              |   21 ---
b/init/do_mounts_initrd.c       |   13 --
b/init/do_mounts_rd.c           |  102 +++++++----------
b/init/initramfs.c              |  103 +++++------------
b/init/main.c                   |   16 +-
include/linux/raid/md_u.h       |   13 --
24 files changed, 251 insertions(+), 452 deletions(-)
Quoted 3 lines
Quoted 1 line

I guess I could say something here... ;)

Like adding an ACK? :)

For the md bits:

Thanks!