Hey, no delays this week.
And nothing really odd going on either. Maybe things are finally calming down.
Sure, I'd have liked it to be even calmer, and there's is movement in
various areas: btrfs updates along with scsi and media driver updates.
And various noise elsewhere. But on the whole it's been pretty boring,
which is just how I like it.
A couple of the fixes were directly linked to my -rc4 announcement,
where I told people they could use a 64-bit kernel to avoid the FP
state save/restore problem we used to have. That brought up some
fairly corner-case compatibility issues that could make that
impractical. So hopefully it's *true* this time around.
And while the FP state save problem is gone, if you have a 64-bit
capable CPU but are still running a 32-bit distro, it really would be
interesting to verify that a 64-bit kernel works for you without
problems. Because it always *should* have, but clearly that wasn't
always the case. It would be very interesting to hear from people who
have perhaps tried and failed before and perhaps didn't even bother to
report the failure? Maybe it's worth trying again?
Linus
---
Al Viro (2):
vfs: fix compat_sys_stat() handling of overflows in st_nlink
ocfs2: deal with wraparounds of i_nlink in ocfs2_rename()
Alan Stern (2):
[SCSI] scsi_pm: Fix bug in the SCSI power management handler
usb-storage: fix freezing of the scanning thread
Alex Deucher (1):
drm/radeon/kms/atom: dpms bios scratch reg updates
Alexander Stein (2):
m68k: Add shared bit to Coldfire kernel page entries
m68k: Do not set global share for non-kernel shared pages
Alexey Khoroshilov (1):
[SCSI] mpt2sas: Fix mismatch in
mpt2sas_base_hard_reset_handler() mutex lock-unlock
Andrew Lunn (1):
USB: Serial: ti_usb_3410_5052: Add Abbot Diabetes Care cable id
Andrew Vasquez (4):
[SCSI] qla2xxx: Add an "is reset active" helper.
[SCSI] qla2xxx: Clear options-flags while issuing stop-firmware
mbx command.
[SCSI] qla2xxx: Remove errant clearing of MBX_INTERRUPT flag
during CT-IOCB processing.
[SCSI] qla2xxx: Correct out of bounds read of ISP2200 mailbox registers.
Anton Altaparmakov (1):
Restore direct_io / truncate locking API
Anton Vorontsov (1):
mm: memcg: Correct unregistring of events attached to the same eventfd
Arne Jansen (1):
btrfs: don't check DUP chunks twice
Arun Easi (1):
[SCSI] qla2xxx: Propagate up abort failures.
Ben Hutchings (1):
builddeb: Don't create files in /tmp with predictable names
Benjamin Herrenschmidt (2):
cpuidle: Default y on powerpc pSeries
powerpc: Fix various issues with return to userspace
Bruno Thomsen (1):
USB: Added Kamstrup VID/PIDs to cp210x serial driver.
Catalin Marinas (1):
ARM: 7323/1: Do not allow ARM_LPAE on pre-ARMv7 architectures
Chad Dupuis (2):
[SCSI] qla2xxx: Add check for null fcport references in
qla2xxx_queuecommand.
[SCSI] qla2xxx: Update version number to 8.03.07.13-k.
Chris D Schimp (2):
hwmon: (max6639) Fix FAN_FROM_REG calculation
hwmon: (max6639) Fix PPR register initialization to set both channels
Chris Mason (6):
Btrfs: don't reserve data with extents locked in btrfs_fallocate
Btrfs: improve error handling for btrfs_insert_dir_item callers
Btrfs: make sure we update latest_bdev
Btrfs: add extra sanity checks on the path names in btrfs_mksubvol
Btrfs: clear the extent uptodate bits during parent transid failures
Btrfs: fix compiler warnings on 32 bit systems
Danny Kukawka (1):
arch/arm/mach-shmobile/board-ag5evm.c: included linux/dma-mapping.h twice
Dave Jiang (1):
[SCSI] isci: Fix NULL ptr dereference when no firmware is being loaded
David Howells (2):
NOMMU: Lock i_mmap_mutex for access to the VMA prio list
NOMMU: Don't need to clear vm_mm when deleting a VMA
David Sterba (2):
btrfs: fix structs where bitfields and spinlock/atomic share 8B word
btrfs: silence warning in raid array setup
Dimitri Sivanich (1):
vfs: fix panic in __d_lookup() with high dentry hashtable counts
Dmitry Kasatkin (1):
digsig: changed type of the timestamp
Doug Ledford (1):
mlx4_core: Exported functions can't be static
Elric Fu (1):
USB: Set hub depth after USB3 hub reset
Eric Paris (1):
ARM/audit: include audit header and fix audit arch
Eugeni Dodonov (4):
drm/i915: gen7: implement rczunit workaround
drm/i915: gen7: Implement an L3 caching workaround.
drm/i915: gen7: work around a system hang on IVB
drm/i915: do not enable RC6p on Sandy Bridge
Florian Albrechtskirchinger (1):
btrfs: honor umask when creating subvol root
Giridhar Malavali (2):
[SCSI] qla2xxx: Complete mailbox command timedout to avoid
initialization failures during next reset cycle.
[SCSI] qla2xxx: Proper detection of firmware abort error code for ISP82xx.
Greg Dietsche (1):
coccicheck: change handling of C={1,2} when M= is set
Greg Ungerer (1):
m68knommu: fix syscall tracing stuck process
Guennadi Liakhovetski (4):
ARM: mach-shmobile: simplify MMCIF DMA configuration
sh: sh7757lcr: update to the new MMCIF DMA configuration
ARM: mach-shmobile: add GPIO-to-IRQ translation to sh7372
sh: se7724: fix compile breakage
Guenter Roeck (2):
hwmon: (ads1015) Fix file leak in probe function
hwmon: (max34440) Fix resetting temperature history
Heiko Carstens (1):
[S390] idle: avoid RCU usage in extended quiescent state
Hitoshi Mitake (1):
asm-generic: architecture independent readq/writeq for 32bit environment
Huajun Li (1):
[SCSI] scsi_scan: Fix 'Poison overwritten' warning caused by
using freed 'shost'
Ian Kent (1):
autofs: work around unhappy compat problem on x86-64
Inki Dae (4):
drm/exynos: added possible_clones setup function.
drm/exynos: fixed page flip issue.
drm/exynos: removed exynos_drm_fbdev_recreate function.
drm/exynos: added postclose to release resource.
James Morris (1):
maintainers: update my email address
Jan Kara (2):
vfs: Provide function to get superblock and wait for it to thaw
quota: Fix deadlock with suspend and quotas
Jan Schmidt (1):
Btrfs: avoid positive number with ERR_PTR
Janne Grunau (1):
[media] hdpvr: fix race conditon during start of streaming
Jarod Wilson (1):
[media] imon: don't wedge hardware after early callbacks
Javi Merino (2):
ARM: 7164/3: PL330: Fix the size of the dst_cache_ctrl field
ARM: 7326/2: PL330: fix null pointer dereference in pl330_chan_ctrl()
Jeff Liu (1):
Btrfs: return the internal error unchanged if
btrfs_get_extent_fiemap() call failed for SEEK_DATA/SEEK_HOLE inquiry
Jeff Mahoney (1):
btrfs: delalloc for page dirtied out-of-band in fixup worker
Jerome Glisse (1):
drm/radeon/kms: properly set accel working flag and bailout when false
Jesper Juhl (1):
XFS: xfs_trans_add_item() - don't assign in ASSERT() when
compare is intended
Joonyoung Shim (2):
drm/exynos: changed priority of mixer layers.
drm/exynos: removed pageflip_event_list init code when closed.
Keith Mannthey (1):
btrfs: Sector Size check during Mount
Kenneth Graunke (1):
drm/i915: gen7: Disable the RHWO optimization as it can cause GPU hangs.
Kleber Sacilotto de Souza (1):
[SCSI] ipr: fix eeh recovery for 64-bit adapters
Kuninori Morimoto (6):
sh: clock-sh7724: fixup sh_fsi clock settings
ARM: mach-shmobile: mackerel: use renesas_usbhs instead of r8a66597_hcd
ARM: mach-shmobile: mackerel: add ak4642 amixer settings on comment
ARM: mach-shmobile: clock-sh73a0: add DSIxPHY clock support
sh: clkfwk: bugfix: use clk_reparent() for div6 clocks
ASoC: ak4642: fixup HeadPhone L/R dapm settings
Kyle McMartin (1):
MAINTAINERS: drop me from PA-RISC maintenance
Laurent Pinchart (1):
[media] omap3isp: Fix crash caused by subdevs now having a
pointer to devnodes
Linus Torvalds (6):
i387: fix up some fpu_counter confusion
i387: use 'restore_fpu_checking()' directly in task switching code
i387: support lazy restore of FPU state
i387: export 'fpu_owner_task' per-cpu variable
sys_poll: fix incorrect type for 'timeout' parameter
Linux 3.3-rc5
Liu Bo (6):
Btrfs: fix trim 0 bytes after a device delete
Btrfs: skip states when they does not contain bits to clear
Btrfs: kick out redundant stuff in convert_extent_bit
Btrfs: fix a bug on overcommit stuff
Btrfs: be less strict on finding next node in clear_extent_bit
Btrfs: increase the global block reserve estimates
Lucas De Marchi (1):
kbuild: do not check for ancient modutils tools
Magnus Damm (5):
ARM: mach-shmobile: sh73a0 PINT IRQ base fix
ARM: mach-shmobile: sh73a0 IRQ sparse alloc fix
ARM: mach-shmobile: IRQ driven GPIO key support for Kota2
ARM: mach-shmobile: sh73a0 PSTR 32-bit access fix
ARM: mach-shmobile: r8a7779 PFC IPSR4 fix
Mark Brown (1):
ASoC: wm8962: Fix sidetone enumeration texts
Mark Hills (1):
ALSA: snd-usb-caiaq: Fix the return of XRUN
Martin Schwidefsky (3):
[S390] incorrect PageTables counter for kvm page tables
[S390] 3215 deadlock with tty_wakeup
[S390] correct ktime to tod clock comparator conversion
Masanari Iida (2):
drm/exynos: Fix typo in exynos_mixer.c
sh: Fix typo in pci-sh7780.c
Miao Xie (1):
Btrfs: fix deadlock on page lock when doing auto-defragment
Michael Christie (1):
[SCSI] qla2xxx: Remove check for null fcport from host reset handler.
Michael Ellerman (1):
powerpc: Fix program check handling when lockdep is enabled
Michel D?nzer (1):
drm/radeon: Only create additional ring debugfs files on Cayman or newer.
Miklos Szeredi (1):
vfs: fix d_inode_lookup() dentry ref leak
Mitsuo Hayasaka (2):
xfs: change available ranges of softlimit and hardlimit in quota check
xfs: make inode quota check more general
Moger, Babu (1):
[SCSI] scsi_dh_rdac: Fix for unbalanced reference count
Myron Stowe (1):
ARM/PCI: Remove ARM's duplicate definition of 'pcibios_max_latency'
Nikolaus Schulz (1):
hwmon: (f75375s) Fix register write order when setting fans to full speed
Oleg Nesterov (2):
epoll: introduce POLLFREE to flush ->signalfd_wqh before kfree()
epoll: ep_unregister_pollwait() can use the freed pwq->whead
Olof Johansson (1):
ARM: 7327/1: need to include asm/system.h in asm/processor.h
Paul Gortmaker (1):
arm: fix compile failure in mach-shmobile/board-ag5evm.c
Paul Mundt (2):
sh: Defer to asm-generic/device.h.
video: pvr2fb: Fix up spurious section mismatch warnings.
Phil Edworthy (1):
sh: Fix sh2a build error for CONFIG_CACHE_WRITETHROUGH
Rabin Vincent (1):
ARM: 7325/1: fix v7 boot with lockdep enabled
Randy Dunlap (1):
[media] wl128x: fix build errors when GPIOLIB is not enabled
Russell King (1):
ARM: OMAP: fix voltage domain build errors with PM_OPP disabled
Rusty Russell (2):
powerpc: Remove references to cpu_*_map
arch/sh: remove references to cpu_*_map.
Santosh Shilimkar (1):
ARM: 7336/1: smp_twd: Don't register CPUFREQ notifiers if local
timers are not initialised
Sarah Sharp (5):
USB: Remove duplicate USB 3.0 hub feature #defines.
xhci: Fix oops caused by more USB2 ports than USB3 ports.
USB: Don't fail USB3 probe on missing legacy PCI IRQ.
USB: Fix handoff when BIOS disables host PCI device.
xhci: Fix encoding for HS bulk/control NAK rate.
Shimoda, Yoshihiro (4):
sh: fix the sh_mmcif_plat_data in board-sh7757lcr
sh: modify resource for SPI0 in setup-sh7757
sh: add platform_device for SPI1 in setup-sh7757
sh: modify a resource of sh_eth_giga1_resources in board-sh7757lcr
Shyam Sundar (1):
[SCSI] qla2xxx: Remove resetting memory during device
initialization for ISP82xx.
Stephen Warren (1):
Kbuild: Use dtc's -d (dependency) option
Steven Rostedt (1):
autofs4 - fix lockdep splat in autofs
Takashi Iwai (2):
ALSA: hda/realtek - Fix overflow of vol/sw check bitmap
ALSA: hda/realtek - Fix surround output regression on Acer Aspire 5935
Taylor Ralph (1):
[media] hdpvr: update picture controls to support firmware versions > 0.15
Tony Lindgren (1):
ARM: 7324/1: modpost: Fix section warnings for ARM for many compilers
Trond Myklebust (3):
NFSv4: Fix an Oops in the NFSv4 getacl code
NFSv4: Ensure we throw out bad delegation stateids on NFS4ERR_BAD_STATEID
NFSv4.1: Fix a NFSv4.1 session initialisation regression
Tsutomu Itoh (3):
Btrfs: fix memory leak in load_free_space_cache()
Btrfs: fix return value check of extent_io_ops
Btrfs: check return value of lookup_extent_mapping() correctly
Weston Andros Adamson (1):
NFSv4: fix server_scope memory leak
Xi Wang (1):
ALSA: usb-audio: avoid integer overflow in create_fixed_stream_quirk()
[email protected] (1):
USB: option: cleanup zte 3g-dongle's pid in option.c
On 12-02-25 03:34 PM, Linus Torvalds wrote:
..
> And while the FP state save problem is gone, if you have a 64-bit
> capable CPU but are still running a 32-bit distro, it really would be
> interesting to verify that a 64-bit kernel works for you without
> problems. Because it always *should* have, but clearly that wasn't
> always the case. It would be very interesting to hear from people who
> have perhaps tried and failed before and perhaps didn't even bother to
> report the failure? Maybe it's worth trying again?
..
The i8k driver is still b0rked for COMPAT use in linux-3.2.xx,
and I don't think my patch got picked up by anyone for 3.3 yet:
(unmangled copy attached -- copy below is for viewing only)
-----snip-----
Fix the i8k (Dell Laptop) driver so that its misdefined ioctl's
work with existing 32-bit userspace on a 64-bit kernel.
Without this patch, 32-bit binaries get EINVAL from most ioctls,
due to incompatible ioctl numbering and lack of a compat_ioctl() function.
The problem is that the ioctls were originally defined using _IOR/_IOW
macros, but the code is written assuming simpler _IO macros. Ugh.
Too late to re-do all of that -- would break existing userspace.
So just fix things up so they work again.
This has been tested with existing pre-compiled binaries (i8kctl)
on both a pure 32-bit system and a 64/32 system.
Signed-off-by: Mark Lord <[email protected]>
--- old/drivers/char/i8k.c 2011-11-28 17:47:43.000000000 -0500
+++ linux/drivers/char/i8k.c 2011-12-10 17:42:58.463023349 -0500
@@ -29,6 +29,7 @@
#include <linux/mutex.h>
#include <linux/hwmon.h>
#include <linux/hwmon-sysfs.h>
+#include <linux/compat.h>
#include <asm/uaccess.h>
#include <asm/io.h>
@@ -92,6 +93,23 @@
static int i8k_open_fs(struct inode *inode, struct file *file);
static long i8k_ioctl(struct file *, unsigned int, unsigned long);
+#ifdef CONFIG_COMPAT
+static long
+i8k_ioctl_unlocked(struct file *fp, unsigned int cmd, int __user *argp);
+
+static long
+i8k_compat_ioctl(struct file *fp, unsigned int cmd, unsigned long arg)
+{
+ long ret;
+
+ mutex_lock(&i8k_mutex);
+ ret = i8k_ioctl_unlocked(fp, cmd, compat_ptr(arg));
+ mutex_unlock(&i8k_mutex);
+
+ return ret;
+}
+#endif
+
static const struct file_operations i8k_fops = {
.owner = THIS_MODULE,
.open = i8k_open_fs,
@@ -99,6 +117,9 @@
.llseek = seq_lseek,
.release = single_release,
.unlocked_ioctl = i8k_ioctl,
+#ifdef CONFIG_COMPAT
+ .compat_ioctl = i8k_compat_ioctl,
+#endif
};
struct smm_regs {
@@ -315,54 +336,56 @@
return regs.eax == 1145651527 && regs.edx == 1145392204 ? 0 : -1;
}
-static int
-i8k_ioctl_unlocked(struct file *fp, unsigned int cmd, unsigned long arg)
+static long
+i8k_ioctl_unlocked(struct file *fp, unsigned int cmd, int __user *argp)
{
int val = 0;
int speed;
unsigned char buff[16];
- int __user *argp = (int __user *)arg;
if (!argp)
return -EINVAL;
- switch (cmd) {
- case I8K_BIOS_VERSION:
- val = i8k_get_bios_version();
+ switch (_IOC_NR(cmd)) {
+ case (_IOC_NR(I8K_BIOS_VERSION)):
+ {
+ int i;
+ for (val = 0, i = 0; i < strlen(bios_version); i++)
+ val = (val << 8) | bios_version[i];
break;
-
- case I8K_MACHINE_ID:
+ }
+ case (_IOC_NR(I8K_MACHINE_ID)):
memset(buff, 0, 16);
strlcpy(buff, i8k_get_dmi_data(DMI_PRODUCT_SERIAL), sizeof(buff));
break;
- case I8K_FN_STATUS:
+ case (_IOC_NR(I8K_FN_STATUS)):
val = i8k_get_fn_status();
break;
- case I8K_POWER_STATUS:
+ case (_IOC_NR(I8K_POWER_STATUS)):
val = i8k_get_power_status();
break;
- case I8K_GET_TEMP:
+ case (_IOC_NR(I8K_GET_TEMP)):
val = i8k_get_temp(0);
break;
- case I8K_GET_SPEED:
+ case (_IOC_NR(I8K_GET_SPEED)):
if (copy_from_user(&val, argp, sizeof(int)))
return -EFAULT;
val = i8k_get_fan_speed(val);
break;
- case I8K_GET_FAN:
+ case (_IOC_NR(I8K_GET_FAN)):
if (copy_from_user(&val, argp, sizeof(int)))
return -EFAULT;
val = i8k_get_fan_status(val);
break;
- case I8K_SET_FAN:
+ case (_IOC_NR(I8K_SET_FAN)):
if (restricted && !capable(CAP_SYS_ADMIN))
return -EPERM;
@@ -383,12 +406,12 @@
return val;
switch (cmd) {
- case I8K_BIOS_VERSION:
+ case (_IOC_NR(I8K_BIOS_VERSION)):
if (copy_to_user(argp, &val, 4))
return -EFAULT;
break;
- case I8K_MACHINE_ID:
+ case (_IOC_NR(I8K_MACHINE_ID)):
if (copy_to_user(argp, buff, 16))
return -EFAULT;
@@ -406,9 +429,10 @@
static long i8k_ioctl(struct file *fp, unsigned int cmd, unsigned long arg)
{
long ret;
+ int __user *argp = (int __user *)arg;
mutex_lock(&i8k_mutex);
- ret = i8k_ioctl_unlocked(fp, cmd, arg);
+ ret = i8k_ioctl_unlocked(fp, cmd, argp);
mutex_unlock(&i8k_mutex);
return ret;
On Sat, Feb 25, 2012 at 12:34:20PM -0800, Linus Torvalds wrote:
> Hey, no delays this week.
>
> And nothing really odd going on either. Maybe things are finally calming down.
>
> Sure, I'd have liked it to be even calmer, and there's is movement in
> various areas: btrfs updates along with scsi and media driver updates.
> And various noise elsewhere. But on the whole it's been pretty boring,
> which is just how I like it.
>
> A couple of the fixes were directly linked to my -rc4 announcement,
> where I told people they could use a 64-bit kernel to avoid the FP
> state save/restore problem we used to have. That brought up some
> fairly corner-case compatibility issues that could make that
> impractical. So hopefully it's *true* this time around.
>
> And while the FP state save problem is gone, if you have a 64-bit
> capable CPU but are still running a 32-bit distro, it really would be
> interesting to verify that a 64-bit kernel works for you without
> problems. Because it always *should* have, but clearly that wasn't
> always the case. It would be very interesting to hear from people who
> have perhaps tried and failed before and perhaps didn't even bother to
> report the failure? Maybe it's worth trying again?
>
> Linus
Yes, I just tried.
IIRC, I used to run 64 bit upstream kernels with 32 bit userspace
under Fedora 14 (because I can) this stopped booting when I switched
to Fedora 15, I think the upstream kernel was 2.6.39 at that time,
so I stopped trying.
With 3.3.0-rc5, a 64 bit kernel works fine for me, again.
Yay!
As a side note, it's a pity that gcc in 32 bit Fedora
does not support -m 64 to support building 64 bit kernels.
--
MST
On Mon, Feb 27, 2012 at 7:23 AM, Mark Lord <[email protected]> wrote:
>
> The i8k driver is still b0rked for COMPAT use in linux-3.2.xx,
> and I don't think my patch got picked up by anyone for 3.3 yet:
I really don't like your patch.
It basically compares the ioctl argument by dropping almost all bits.
That looks completely wrong. If I read the patch correctly, it will
now match ioctl's that have *nothing* to do with the i8k ioctls, that
just happen to have a few bits in common.
I also think that your explanation is wrong. The problem is not the
use of _IOR() at all, the problem is that the data structure given
*to* the _IOR() macro is complete grabage. For example:
#define I8K_GET_SPEED _IOWR('i', 0x85, size_t)
that "size_t" there is just crap. It's wrong. The ioctl doesn't write
a size_t (which is different sizes on 32-bit and 64-bit), it actually
writes an "int" (which happens to be the same size in both compat and
non-compat).
So the 64-bit code - totally incorrectly - has been compiled with an
ioctl number that implies a 64-bit argument.
EVERY SINGLE IOCTL that is defined for the i8k driver is crap. There
are two that were "int", but they are marked as "broken" because the
data structure *they* write isn't actually "int". But those two are
the ones that just happen to work in both 32-bit and 64-bit mode,
because at least the size of the (incorrect) data structure is the
same.
The ones that have "size_t" aren't marked broken, but those are the
*really* broken ones due to the wrong data structure choice that has a
size that now is different in 32-bit and 64-bit mode.
Quite frankly, I think the right solution is to fix the kernel
interface to the right type (int) that is the same. But because we
don't want to change the user interface, let's make the kernel
*accept* the 8-byte entity and just change it into a 4-byte size, and
leave the user-space visible ioctl numbers the same broken crap - it's
not like the other ioctl numbers had the right size *either*...
IOW, have something like the attached in the ioctl handler (and then
we need to also add the compat handler, as in your patch).
Does this (with your compat_ioctl wrapper addition) also work for you?
Linus
Linus Torvalds <[email protected]> writes:
> I also think that your explanation is wrong. The problem is not the
> use of _IOR() at all, the problem is that the data structure given
> *to* the _IOR() macro is complete grabage. For example:
>
> #define I8K_GET_SPEED _IOWR('i', 0x85, size_t)
Originally the garbage looked like this:
#define I8K_BIOS_VERSION _IOR ('i', 0x80, 4)
#define I8K_MACHINE_ID _IOR ('i', 0x81, 16)
#define I8K_POWER_STATUS _IOR ('i', 0x82, sizeof(int))
#define I8K_FN_STATUS _IOR ('i', 0x83, sizeof(int))
#define I8K_GET_TEMP _IOR ('i', 0x84, sizeof(int))
#define I8K_GET_SPEED _IOWR('i', 0x85, sizeof(int))
#define I8K_GET_FAN _IOWR('i', 0x86, sizeof(int))
#define I8K_SET_FAN _IOWR('i', 0x87, sizeof(int)*2)
Later they were modified by "[PATCH] use size_t for the broken ioctl
numbers" and "Fix more ioctl _IOR/_IOW misusage." (see tglx's history
tree).
Andreas.
--
Andreas Schwab, [email protected]
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5
"And now for something completely different."
On Tue, Feb 28, 2012 at 11:29 AM, Andreas Schwab <[email protected]> wrote:
>
> Originally the garbage looked like this:
>
> #define I8K_BIOS_VERSION ? ? ? _IOR ('i', 0x80, 4)
> #define I8K_MACHINE_ID ? ? ? ? _IOR ('i', 0x81, 16)
> #define I8K_POWER_STATUS ? ? ? _IOR ('i', 0x82, sizeof(int))
> #define I8K_FN_STATUS ? ? ? ? ?_IOR ('i', 0x83, sizeof(int))
> #define I8K_GET_TEMP ? ? ? ? ? _IOR ('i', 0x84, sizeof(int))
> #define I8K_GET_SPEED ? ? ? ? ?_IOWR('i', 0x85, sizeof(int))
> #define I8K_GET_FAN ? ? ? ? ? ?_IOWR('i', 0x86, sizeof(int))
> #define I8K_SET_FAN ? ? ? ? ? ?_IOWR('i', 0x87, sizeof(int)*2)
>
> Later they were modified by "[PATCH] use size_t for the broken ioctl
> numbers" and "Fix more ioctl _IOR/_IOW misusage." (see tglx's history
> tree).
Ahh, yes. That wasn't the only driver that did the "sizeof" by hand,
and as a result got "size_t" when the (nested) *real* sizeof then
happened.
That explains the history behind the garbage.
ioctl numbering have always been pretty painful. Oh, how I wish they
had been ASCII strings rather than random binary blobs. But that's the
way things are (and we certainly screwed up a few /proc files too, so
it's not like using text fixes all potential for screwing up)
Linus