14 years ago[IA64] ptrace and restore_sigcontext() allow ar.rsc.pl==0
Matthew Chapman [Tue, 21 Jun 2005 23:19:20 +0000]
[IA64] ptrace and restore_sigcontext() allow ar.rsc.pl==0

This patch fixes handling of accesses to ar.rsc via ptrace & restore_sigcontext
[With Thanks to Chris Wright for noticing the restore_sigcontext path]

Signed-off-by: Matthew Chapman <matthewc@hp.com>
Acked-by: David Mosberger <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

14 years ago[IA64] remove "pci=routeirq" option
Bjorn Helgaas [Mon, 13 Jun 2005 18:27:00 +0000]
[IA64] remove "pci=routeirq" option

Remove "pci=routeirq" option for ia64.  This was a workaround
after ACPI IRQ routing was changed from "all at boot for everything
in _PRT" to "do it when the device is enabled" in case there were
drivers that didn't use pci_enable_device().

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

14 years ago[IA64] fix nested_dtlb_miss handler for hugetlb address
Ken Chen [Tue, 21 Jun 2005 21:40:31 +0000]
[IA64] fix nested_dtlb_miss handler for hugetlb address

The nested_dtlb_miss handler currently does not handle fault from
hugetlb address correctly.  It walks the page table assuming PAGE_SIZE.
Thus when taking a fault triggered from hugetlb address, it would not
calculate the pgd/pmd/pte address correctly and thus result an incorrect
invocation of ia64_do_page_fault().  In there, kernel will signal SIGBUS
and application dies (The faulting address is perfectly legal and we
have a valid pte for the corresponding user hugetlb address as well).
This patch fix the described kernel bug.  Since nested_dtlb_miss is a
rare event and a slow path anyway, I'm making the change without #ifdef
CONFIG_HUGETLB_PAGE for code readability.  Tony, please apply.

Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

14 years ago[IA64] printk needs KERN_INFO arch/ia64/kernel/smp.c
Christophe Lucas [Mon, 20 Jun 2005 21:50:00 +0000]
[IA64] printk needs KERN_INFO arch/ia64/kernel/smp.c

printk() calls should include appropriate KERN_* constant.

Signed-off-by: Christophe Lucas <clucas@rotomalug.org>
Signed-off-by: Domen Puncer <domen@coderock.org>
Signed-off-by: Tony Luck <tony.luck@intel.com>

14 years ago[IA64] enable SGI simulator for generic kernels
Greg Edwards [Tue, 21 Jun 2005 18:30:00 +0000]
[IA64] enable SGI simulator for generic kernels

Allow the SGI simulator (medusa) to work on generic kernels.  There is
no inherent dependency on an sn2-specific kernel.

Boot tested on Altix, medusa and HP rx2600.

Signed-off-by: Greg Edwards <edwardsg@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

14 years ago[IA64] Refresh tiger_defconfig
Tony Luck [Tue, 21 Jun 2005 21:16:00 +0000]
[IA64] Refresh tiger_defconfig

Signed-off-by: Tony Luck <tony.luck@intel.com>

14 years ago[IA64] refresh arch/ia64/defconfig
Greg Edwards [Tue, 21 Jun 2005 17:43:00 +0000]
[IA64] refresh arch/ia64/defconfig

Refresh arch/ia64/defconfig, as it was getting a bit stale.  The only
manual changes I made were:

CONFIG_SCSI_SATA=y needed for some Altix base I/O cards
CONFIG_SCSI_SATA_VITESSE=y

CONFIG_DM_MULTIPATH=m the rest are already modules

CONFIG_FUSION_SPI=y new driver breakout
CONFIG_FUSION_FC=m

CONFIG_SGI_TIOCX=y enable some other SGI drivers
CONFIG_SGI_MBCS=m
CONFIG_AGP_SGI_TIOCA=m

Boot tested on Altix, HP rx2600 and Intel Tiger

Signed-off-by: Greg Edwards <edwardsg@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

14 years agoAuto merge with /home/aegl/GIT/linus
Tony Luck [Mon, 20 Jun 2005 16:35:34 +0000]
Auto merge with /home/aegl/GIT/linus

14 years ago[IA64] Drop spurious paren in entry.h
David Mosberger-Tang [Fri, 10 Jun 2005 05:36:00 +0000]
[IA64] Drop spurious paren in entry.h

The latest assembler catches this typo.  (reported by Jim Wilson).

Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

14 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Sun, 19 Jun 2005 18:53:06 +0000]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

14 years agoMerge with ../linux-2.6-smp
Russell King [Sun, 19 Jun 2005 18:26:54 +0000]
Merge with ../linux-2.6-smp

14 years ago[PATCH] ARM SMP: Fix PXA/SA11x0 suspend resume crash
Russell King [Sun, 19 Jun 2005 17:39:33 +0000]
[PATCH] ARM SMP: Fix PXA/SA11x0 suspend resume crash

We need to re-initialise the stack pointers for undefined, IRQ
and abort mode handlers whenever we resume.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

14 years ago[PATCH] ARM SMP: Add missed files from Integrator/CP platform
Russell King [Sun, 19 Jun 2005 08:52:07 +0000]
[PATCH] ARM SMP: Add missed files from Integrator/CP platform

Add missed new files from basic SMP support for the Integrator/CP platform.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

14 years ago[TCP]: Fix sysctl_tcp_low_latency
David S. Miller [Sun, 19 Jun 2005 06:01:10 +0000]
[TCP]: Fix sysctl_tcp_low_latency

When enabled, this should disable UCOPY prequeue'ing altogether,
but it does not due to a missing test.

Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[IPV4]: [4/4] signed vs unsigned cleanup in net/ipv4/raw.c
Jesper Juhl [Sun, 19 Jun 2005 06:00:34 +0000]
[IPV4]: [4/4] signed vs unsigned cleanup in net/ipv4/raw.c

This patch changes the type of the third parameter 'length' of the
raw_send_hdrinc() function from 'int' to 'size_t'.
This makes sense since this function is only ever called from one
location, and the value passed as the third parameter in that location is
itself of type size_t, so this makes the recieving functions parameter
type match. Also, inside raw_send_hdrinc() the 'length' variable is
used in comparisons with unsigned values and passed as parameter to
functions expecting unsigned values (it's used in a single comparison with
a signed value, but that one can never actually be negative so the patch
also casts that one to size_t to stop gcc worrying, and it is passed in a
single instance to memcpy_fromiovecend() which expects a signed int, but
as far as I can see that's not a problem since the value of 'length'
shouldn't ever exceed the value of a signed int).

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[IPV4]: [3/4] signed vs unsigned cleanup in net/ipv4/raw.c
Jesper Juhl [Sun, 19 Jun 2005 06:00:15 +0000]
[IPV4]: [3/4] signed vs unsigned cleanup in net/ipv4/raw.c

This patch changes the type of the local variable 'i' in
raw_probe_proto_opt() from 'int' to 'unsigned int'. The only use of 'i' in
this function is as a counter in a for() loop and subsequent index into
the msg->msg_iov[] array.
Since 'i' is compared in a loop to the unsigned variable msg->msg_iovlen
gcc -W generates this warning :

net/ipv4/raw.c:340: warning: comparison between signed and unsigned

Changing 'i' to unsigned silences this warning and is safe since the array
index can never be negative anyway, so unsigned int is the logical type to
use for 'i' and also enables a larger msg_iov[] array (but I don't know if
that will ever matter).

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[IPV4]: [2/4] signed vs unsigned cleanup in net/ipv4/raw.c
Jesper Juhl [Sun, 19 Jun 2005 06:00:00 +0000]
[IPV4]: [2/4] signed vs unsigned cleanup in net/ipv4/raw.c

This patch gets rid of the following gcc -W warning in net/ipv4/raw.c :

net/ipv4/raw.c:387: warning: comparison of unsigned expression < 0 is always false

Since 'len' is of type size_t it is unsigned and can thus never be <0, and
since this is obvious from the function declaration just a few lines above
I think it's ok to remove the pointless check for len<0.

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[IPV4]: [1/4] signed vs unsigned cleanup in net/ipv4/raw.c
Jesper Juhl [Sun, 19 Jun 2005 05:59:45 +0000]
[IPV4]: [1/4] signed vs unsigned cleanup in net/ipv4/raw.c

This patch silences these two gcc -W warnings in net/ipv4/raw.c :

net/ipv4/raw.c:517: warning: signed and unsigned type in conditional expression
net/ipv4/raw.c:613: warning: signed and unsigned type in conditional expression

It doesn't change the behaviour of the code, simply writes the conditional
expression with plain 'if()' syntax instead of '? :' , but since this
breaks it into sepperate statements gcc no longer complains about having
both a signed and unsigned value in the same conditional expression.

Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: noop/noqueue qdisc style cleanups
Thomas Graf [Sun, 19 Jun 2005 05:59:08 +0000]
[PKT_SCHED]: noop/noqueue qdisc style cleanups

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: Cleanup pfifo_fast qdisc and remove unnecessary code
Thomas Graf [Sun, 19 Jun 2005 05:58:53 +0000]
[PKT_SCHED]: Cleanup pfifo_fast qdisc and remove unnecessary code

Removes the skb trimming code which is not needed since we never
touch the skb upon failure. Removes unnecessary initializers,
and simplifies the code a bit.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: Add and use prio2list() in the pfifo_fast qdisc
Thomas Graf [Sun, 19 Jun 2005 05:58:35 +0000]
[PKT_SCHED]: Add and use prio2list() in the pfifo_fast qdisc

prio2list() returns the relevant sk_buff_head for the
band specified by the priority for a given skb.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: Transform pfifo_fast to use generic queue management interface
Thomas Graf [Sun, 19 Jun 2005 05:58:15 +0000]
[PKT_SCHED]: Transform pfifo_fast to use generic queue management interface

Gives pfifo_fast a byte based backlog.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: Cleanup fifo qdisc and remove unnecessary code
Thomas Graf [Sun, 19 Jun 2005 05:58:00 +0000]
[PKT_SCHED]: Cleanup fifo qdisc and remove unnecessary code

Removes the skb trimming code which is not needed since we never
touch the skb upon failure. Removes unnecessary includes,
initializers, and simplifies the code a bit.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: Transform fifo qdisc to use generic queue management interface
Thomas Graf [Sun, 19 Jun 2005 05:57:42 +0000]
[PKT_SCHED]: Transform fifo qdisc to use generic queue management interface

The simplicity of the fifo qdisc allows several qdisc operations to be
redirected to the relevant queue management function directly. Saves
a lot of code lines and gives the pfifo a byte based backlog.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: Generic queue management interface for qdiscs using internal skb queues
Thomas Graf [Sun, 19 Jun 2005 05:57:26 +0000]
[PKT_SCHED]: Generic queue management interface for qdiscs using internal skb queues

Implements an interface to be used by leaf qdiscs maintaining an internal
skb queue. The interface maintains a backlog in bytes additionaly
to the skb_queue_len() maintained by the queue itself. Relevant statistics
get incremented automatically. Every function comes in two variants, one
assuming Qdisc->q is used as queue and the second taking a sk_buff_head
as argument. Be aware that, if you use multiple queues, you still have to
maintain the Qdisc->q.qlen counter yourself.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[SCTP]: Replace spin_lock_irqsave with spin_lock_bh
Herbert Xu [Sun, 19 Jun 2005 05:56:42 +0000]
[SCTP]: Replace spin_lock_irqsave with spin_lock_bh

This patch replaces the spin_lock_irqsave call on the receive queue
lock in SCTP with spin_lock_bh.  Despite the proliferation of
spin_lock_irqsave calls in this stack, it is only entered from the
IPv4/IPv6 stack and user space.  That is, it is never entered from
hardirq context.

The call in question is only called from recvmsg which means that
IRQs aren't disabled.  Therefore it is safe to replace it with
spin_lock_bh.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[IPV4/IPV6]: Replace spin_lock_irq with spin_lock_bh
Herbert Xu [Sun, 19 Jun 2005 05:56:18 +0000]
[IPV4/IPV6]: Replace spin_lock_irq with spin_lock_bh

In light of my recent patch to net/ipv4/udp.c that replaced the
spin_lock_irq calls on the receive queue lock with spin_lock_bh,
here is a similar patch for all other occurences of spin_lock_irq
on receive/error queue locks in IPv4 and IPv6.

In these stacks, we know that they can only be entered from user
or softirq context.  Therefore it's safe to disable BH only.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NETLINK]: Set correct pid for ioctl originating netlink events
Jamal Hadi Salim [Sun, 19 Jun 2005 05:55:51 +0000]
[NETLINK]: Set correct pid for ioctl originating netlink events

This patch ensures that netlink events created as a result of programns
using ioctls (such as ifconfig, route etc) contains the correct PID of
those events.

Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NETLINK]: Explicit typing
Jamal Hadi Salim [Sun, 19 Jun 2005 05:55:31 +0000]
[NETLINK]: Explicit typing

This patch converts "unsigned flags" to use more explict types like u16
instead and incrementally introduces NLMSG_NEW().

Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[DECNET]: Remove unnecessary initilization of unused variable entries
Thomas Graf [Sun, 19 Jun 2005 05:55:02 +0000]
[DECNET]: Remove unnecessary initilization of unused variable entries

This patch was supposed to be part of the neighbour tables related
patchset but apparently got lost.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[IPSEC]: Add XFRMA_SA/XFRMA_POLICY for delete notification
Herbert Xu [Sun, 19 Jun 2005 05:54:36 +0000]
[IPSEC]: Add XFRMA_SA/XFRMA_POLICY for delete notification

This patch changes the format of the XFRM_MSG_DELSA and
XFRM_MSG_DELPOLICY notification so that the main message
sent is of the same format as that received by the kernel
if the original message was via netlink.  This also means
that we won't lose the byid information carried in km_event.

Since this user interface is introduced by Jamal's patch
we can still afford to change it.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NETLINK]: Correctly set NLM_F_MULTI without checking the pid
Jamal Hadi Salim [Sun, 19 Jun 2005 05:54:12 +0000]
[NETLINK]: Correctly set NLM_F_MULTI without checking the pid

This patch rectifies some rtnetlink message builders that derive the
flags from the pid. It is now explicit like the other cases
which get it right. Also fixes half a dozen dumpers which did not
set NLM_F_MULTI at all.

Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NETLINK]: Introduce NLMSG_NEW macro to better handle netlink flags
Thomas Graf [Sun, 19 Jun 2005 05:53:48 +0000]
[NETLINK]: Introduce NLMSG_NEW macro to better handle netlink flags

Introduces a new macro NLMSG_NEW which extends NLMSG_PUT but takes
a flags argument. NLMSG_PUT stays there for compatibility but now
calls NLMSG_NEW with flags == 0. NLMSG_PUT_ANSWER is renamed to
NLMSG_NEW_ANSWER which now also takes a flags argument.

Also converts the users of NLMSG_PUT_ANSWER to use NLMSG_NEW_ANSWER
and fixes the two direct users of __nlmsg_put to either provide
the flags or use NLMSG_NEW(_ANSWER).

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: Logic simplifications and codingstyle/whitespace cleanups
Thomas Graf [Sun, 19 Jun 2005 05:53:29 +0000]
[PKT_SCHED]: Logic simplifications and codingstyle/whitespace cleanups

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: Make dsmark use the new dumping macros
Thomas Graf [Sun, 19 Jun 2005 05:53:12 +0000]
[PKT_SCHED]: Make dsmark use the new dumping macros

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[PKT_SCHED]: Fix dsmark to apply changes consistent
Thomas Graf [Sun, 19 Jun 2005 05:52:54 +0000]
[PKT_SCHED]: Fix dsmark to apply changes consistent

Fixes dsmark to do all configuration sanity checks first and
only apply the changes if all of them can be applied without
any errors. Also fixes the weak sanity checks for DSMARK_VALUE
and DSMASK_MASK.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[RTNETLINK]: Add RTA_(PUT|GET) shortcuts for u8, u16, and flag
Thomas Graf [Sun, 19 Jun 2005 05:52:36 +0000]
[RTNETLINK]: Add RTA_(PUT|GET) shortcuts for u8, u16, and flag

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NEIGH]: Fix use of uninitialized variable when trimming in neightbl_fill_parms
Thomas Graf [Sun, 19 Jun 2005 05:52:09 +0000]
[NEIGH]: Fix use of uninitialized variable when trimming in neightbl_fill_parms

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NETLINK]: Kill bogus NLMSG_SET_MULTIPART uses.
Thomas Graf [Sun, 19 Jun 2005 05:51:43 +0000]
[NETLINK]: Kill bogus NLMSG_SET_MULTIPART uses.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NETLINK]: Fix RTA_NEST_CANCEL().
Thomas Graf [Sun, 19 Jun 2005 05:51:26 +0000]
[NETLINK]: Fix RTA_NEST_CANCEL().

Only skb_trim() if 'start' is non-NULL.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NEIGHBOUR]: Remove unused fields in struct neigh_parms and neigh_table
Thomas Graf [Sun, 19 Jun 2005 05:51:12 +0000]
[NEIGHBOUR]: Remove unused fields in struct neigh_parms and neigh_table

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NETLINK]: Neighbour table configuration and statistics via rtnetlink
Thomas Graf [Sun, 19 Jun 2005 05:50:55 +0000]
[NETLINK]: Neighbour table configuration and statistics via rtnetlink

To retrieve the neighbour tables send RTM_GETNEIGHTBL with the
NLM_F_DUMP flag set. Every neighbour table configuration is
spread over multiple messages to avoid running into message
size limits on systems with many interfaces. The first message
in the sequence transports all not device specific data such as
statistics, configuration, and the default parameter set.
This message is followed by 0..n messages carrying device
specific parameter sets.

Although the ordering should be sufficient, NDTA_NAME can be
used to identify sequences. The initial message can be identified
by checking for NDTA_CONFIG. The device specific messages do
not contain this TLV but have NDTPA_IFINDEX set to the
corresponding interface index.

To change neighbour table attributes, send RTM_SETNEIGHTBL
with NDTA_NAME set. Changeable attribute include NDTA_THRESH[1-3],
NDTA_GC_INTERVAL, and all TLVs in NDTA_PARMS unless marked
otherwise. Device specific parameter sets can be changed by
setting NDTPA_IFINDEX to the interface index of the corresponding
device.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NETLINK] Routing attribute related shortcuts
Thomas Graf [Sun, 19 Jun 2005 05:50:38 +0000]
[NETLINK] Routing attribute related shortcuts

 RTA_GET_U(32|64)(tlv)
   Assumes TLV is a u32/u64 field and returns its value.

 RTA_GET_[M]SECS(tlv)
   Assumes TLV is a u64 and transports jiffies converted
   to seconds or milliseconds and returns its value.

 RTA_PUT_U(32|64)(skb, type, value)
   Appends %value as fixed u32/u64 to %skb as TLV %type.

 RTA_PUT_[M]SECS(skb, type, jiffies)
   Converts %jiffies to secs/msecs and appends it as u64
   to %skb as TLV %type.

 RTA_PUT_STRING(skb, type, string)
   Appends %NUL terminated %string to %skb as TLV %type.

 RTA_NEST(skb, type)
   Starts a nested TLV %type and returns the nesting handle.

 RTA_NEST_END(skb, nesting_handle)
   Finishes the nested TLV %nesting_handle, must be called
   symmetric to RTA_NEST(). Returns skb->len

 RTA_NEST_CANCEL(skb, nesting_handle)
   Cancel the nested TLV %nesting_handle and trim nested TLV
   from skb again, returns -1.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NETLINK]: New message building macros
Thomas Graf [Sun, 19 Jun 2005 05:50:12 +0000]
[NETLINK]: New message building macros

 NLMSG_PUT_ANSWER(skb, nlcb, type, length)
   Start a new netlink message as answer to a request,
   returns the message header.

 NLMSG_END(skb, nlh)
   End a netlink message, fixes total message length,
   returns skb->len.

 NLMSG_CANCEL(skb, nlh)
   Cancel the building process and trim whole message
   from skb again, returns -1.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NET]: Move sysctl_max_syn_backlog into request_sock.c
David S. Miller [Sun, 19 Jun 2005 05:49:40 +0000]
[NET]: Move sysctl_max_syn_backlog into request_sock.c

This fixes the CONFIG_INET=n build failure noticed
by Andrew Morton.

Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NET] rename struct tcp_listen_opt to struct listen_sock
Arnaldo Carvalho de Melo [Sun, 19 Jun 2005 05:48:55 +0000]
[NET] rename struct tcp_listen_opt to struct listen_sock

Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NET] Generalise tcp_listen_opt
Arnaldo Carvalho de Melo [Sun, 19 Jun 2005 05:47:59 +0000]
[NET] Generalise tcp_listen_opt

This chunks out the accept_queue and tcp_listen_opt code and moves
them to net/core/request_sock.c and include/net/request_sock.h, to
make it useful for other transport protocols, DCCP being the first one
to use it.

Next patches will rename tcp_listen_opt to accept_sock and remove the
inline tcp functions that just call a reqsk_queue_ function.

Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NET] Rename open_request to request_sock
Arnaldo Carvalho de Melo [Sun, 19 Jun 2005 05:47:21 +0000]
[NET] Rename open_request to request_sock

Ok, this one just renames some stuff to have a better namespace and to
dissassociate it from TCP:

struct open_request  -> struct request_sock
tcp_openreq_alloc    -> reqsk_alloc
tcp_openreq_free     -> reqsk_free
tcp_openreq_fastfree -> __reqsk_free

With this most of the infrastructure closely resembles a struct
sock methods subset.

Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[NET] Generalise TCP's struct open_request minisock infrastructure
Arnaldo Carvalho de Melo [Sun, 19 Jun 2005 05:46:52 +0000]
[NET] Generalise TCP's struct open_request minisock infrastructure

Kept this first changeset minimal, without changing existing names to
ease peer review.

Basicaly tcp_openreq_alloc now receives the or_calltable, that in turn
has two new members:

->slab, that replaces tcp_openreq_cachep
->obj_size, to inform the size of the openreq descendant for
  a specific protocol

The protocol specific fields in struct open_request were moved to a
class hierarchy, with the things that are common to all connection
oriented PF_INET protocols in struct inet_request_sock, the TCP ones
in tcp_request_sock, that is an inet_request_sock, that is an
open_request.

I.e. this uses the same approach used for the struct sock class
hierarchy, with sk_prot indicating if the protocol wants to use the
open_request infrastructure by filling in sk_prot->rsk_prot with an
or_calltable.

Results? Performance is improved and TCP v4 now uses only 64 bytes per
open request minisock, down from 96 without this patch :-)

Next changeset will rename some of the structs, fields and functions
mentioned above, struct or_calltable is way unclear, better name it
struct request_sock_ops, s/struct open_request/struct request_sock/g,
etc.

Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[SLAB] Introduce kmem_cache_name
Arnaldo Carvalho de Melo [Sun, 19 Jun 2005 05:46:19 +0000]
[SLAB] Introduce kmem_cache_name

This is for use with slab users that pass a dynamically allocated slab name in
kmem_cache_create, so that before destroying the slab one can retrieve the name
and free its memory.

Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[IPSEC] Use NLMSG_LENGTH in xfrm_exp_state_notify
Jamal Hadi Salim [Sun, 19 Jun 2005 05:45:56 +0000]
[IPSEC] Use NLMSG_LENGTH in xfrm_exp_state_notify

Small fixup to use netlink macros instead of hardcoding.

Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[IPSEC] Fix xfrm_state leaks in error path
Patrick McHardy [Sun, 19 Jun 2005 05:45:31 +0000]
[IPSEC] Fix xfrm_state leaks in error path

Herbert Xu wrote:
> @@ -1254,6 +1326,7 @@ static int pfkey_add(struct sock *sk, st
>       if (IS_ERR(x))
>               return PTR_ERR(x);
>
> +     xfrm_state_hold(x);

This introduces a leak when xfrm_state_add()/xfrm_state_update()
fail. We hold two references (one from xfrm_state_alloc(), one
from xfrm_state_hold()), but only drop one. We need to take the
reference because the reference from xfrm_state_alloc() can
be dropped by __xfrm_state_delete(), so the fix is to drop both
references on error. Same problem in xfrm_user.c.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 years ago[IPSEC] Use XFRM_MSG_* instead of XFRM_SAP_*
Herbert Xu [Sun, 19 Jun 2005 05:44:37 +0000]
[IPSEC] Use XFRM_MSG_* instead of XFRM_SAP_*

This patch removes XFRM_SAP_* and converts them over to XFRM_MSG_*.
The netlink interface is meant to map directly onto the underlying
xfrm subsystem.  Therefore rather than using a new independent
representation for the events we can simply use the existing ones
from xfrm_user.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

14 years ago[IPSEC] Set byid for km_event in xfrm_get_policy
Herbert Xu [Sun, 19 Jun 2005 05:44:18 +0000]
[IPSEC] Set byid for km_event in xfrm_get_policy

This patch fixes policy deletion in xfrm_user so that it sets
km_event.data.byid.  This puts xfrm_user on par with what af_key
does in this case.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

14 years ago[IPSEC] Turn km_event.data into a union
Herbert Xu [Sun, 19 Jun 2005 05:44:00 +0000]
[IPSEC] Turn km_event.data into a union

This patch turns km_event.data into a union.  This makes code that
uses it clearer.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

14 years ago[IPSEC] Fix xfrm to pfkey SA state conversion
Herbert Xu [Sun, 19 Jun 2005 05:43:43 +0000]
[IPSEC] Fix xfrm to pfkey SA state conversion

This patch adjusts the SA state conversion in af_key such that
XFRM_STATE_ERROR/XFRM_STATE_DEAD will be converted to SADB_STATE_DEAD
instead of SADB_STATE_DYING.

According to RFC 2367, SADB_STATE_DYING SAs can be turned into
mature ones through updating their lifetime settings.  Since SAs
which are in the states XFRM_STATE_ERROR/XFRM_STATE_DEAD cannot
be resurrected, this value is unsuitable.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

14 years ago[IPSEC] Kill spurious hard expire messages
Herbert Xu [Sun, 19 Jun 2005 05:43:22 +0000]
[IPSEC] Kill spurious hard expire messages

This patch ensures that the hard state/policy expire notifications are
only sent when the state/policy is successfully removed from their
respective tables.

As it is, it's possible for a state/policy to both expire through
reaching a hard limit, as well as being deleted by the user.

Note that this behaviour isn't actually forbidden by RFC 2367.
However, it is a quality of implementation issue.

As an added bonus, the restructuring in this patch will help
eventually in moving the expire notifications from softirq
context into process context, thus improving their reliability.

One important side-effect from this change is that SAs reaching
their hard byte/packet limits are now deleted immediately, just
like SAs that have reached their hard time limits.

Previously they were announced immediately but only deleted after
30 seconds.

This is bad because it prevents the system from issuing an ACQUIRE
command until the existing state was deleted by the user or expires
after the time is up.

In the scenario where the expire notification was lost this introduces
a 30 second delay into the system for no good reason.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

14 years ago[IPSEC] Add complete xfrm event notification
Jamal Hadi Salim [Sun, 19 Jun 2005 05:42:13 +0000]
[IPSEC] Add complete xfrm event notification

Heres the final patch.
What this patch provides

- netlink xfrm events
- ability to have events generated by netlink propagated to pfkey
  and vice versa.
- fixes the acquire lets-be-happy-with-one-success issue

Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

14 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Sat, 18 Jun 2005 23:06:50 +0000]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

14 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/dwmw2/audit-2.6
Linus Torvalds [Sat, 18 Jun 2005 20:54:12 +0000]
Merge /pub/scm/linux/kernel/git/dwmw2/audit-2.6

14 years ago[PATCH] DocBook: MAINTAINER entry
Martin Waitz [Fri, 17 Jun 2005 20:20:59 +0000]
[PATCH] DocBook: MAINTAINER entry

Martin can maintain the DocBook system for us.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

14 years ago[PATCH] console blanking oops fix
Jan Beulich [Fri, 17 Jun 2005 20:20:58 +0000]
[PATCH] console blanking oops fix

When significant delays happen during boot (e.g.  with a kernel debugger,
but the problem has also seen in other cases) the timeout for blanking the
console may trigger, but the work scheduler may not have been initialized,
yet.  schedule_work() will oops over the null keventd_wq.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

14 years agoMerge 'upstream-2.6.13' branch of rsync://rsync.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 18 Jun 2005 20:08:39 +0000]
Merge 'upstream-2.6.13' branch of /linux/kernel/git/jgarzik/libata-dev

14 years agoClean up subthread exec
Linus Torvalds [Sat, 18 Jun 2005 20:06:22 +0000]
Clean up subthread exec

Make sure we re-parent itimers, and use BUG_ON() instead of an explicit
conditional BUG().

14 years ago[PATCH] Fix typo in drivers/pci/pci-driver.c
Mika Kukkonen [Sat, 18 Jun 2005 19:49:56 +0000]
[PATCH] Fix typo in drivers/pci/pci-driver.c

The git commit 794f5bfa77955c4455f6d72d8b0e2bee25f1ff0c
accidentally suffers from a previous typo in that file
(',' instead of ';' in end of line). Patch included.

Signed-off-by: Mika Kukkonen (mikukkon@iki.fi)
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

14 years agoManual merge of rsync://rsync.kernel.org/pub/scm/linux/kernel/git/jgarzik/netdev...
Linus Torvalds [Sat, 18 Jun 2005 18:42:35 +0000]
Manual merge of /linux/kernel/git/jgarzik/netdev-2.6.git

This is a fixed-up version of the broken "upstream-2.6.13" branch, where
I re-did the manual merge of drivers/net/r8169.c by hand, and made sure
the history is all good.

14 years agoMerge /spare/repo/linux-2.6/
Jeff Garzik [Sat, 18 Jun 2005 17:21:24 +0000]
Merge /spare/repo/linux-2.6/

14 years ago[PATCH] ARM SMP: Add basic support Integrator/CP platform
Russell King [Sat, 18 Jun 2005 09:15:46 +0000]
[PATCH] ARM SMP: Add basic support Integrator/CP platform

Add basic SMP support for the Integrator/CP platform.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

14 years ago[PATCH] ARM SMP: Add support for startup of secondary processors
Russell King [Sat, 18 Jun 2005 08:33:31 +0000]
[PATCH] ARM SMP: Add support for startup of secondary processors

Create a temporary page table to startup secondary processors.  This
page table must have a 1:1 virtual/physical mapping for the kernel
in addition to the standard mappings to ensure that the secondary
CPU can enable its MMU safely.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

14 years agoMerge with ../linux-2.6-smp
Russell King [Sat, 18 Jun 2005 08:06:59 +0000]
Merge with ../linux-2.6-smp

14 years agoMerge with master.kernel.org:/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
David Woodhouse [Sat, 18 Jun 2005 07:36:46 +0000]
Merge ... /linux/kernel/git/torvalds/linux-2.6.git

14 years ago[SCSI] Add DMA mask constants other than 32 and 64 bit
Lee Revell [Wed, 15 Jun 2005 18:19:03 +0000]
[SCSI] Add DMA mask constants other than 32 and 64 bit

Signed-Off-By: Lee Revell <rlrevell@joe-job.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

14 years ago[SCSI] aacraid: regression fix
Mark Haverkamp [Fri, 17 Jun 2005 20:38:04 +0000]
[SCSI] aacraid: regression fix

The fixes for sparse warnings mixed in with the fixups for
the raw_srb handler resulted in a bug that showed up in the 32 bit
environments when trying to issue calls directly to the physical devices
that are part of the arrays (ioctl scsi passthrough).

Received from Mark Salyzyn at adaptec.
Applied comment from Christoph to remove cpu_to_le32(0)
Applied Mark S fix of missing memcpy.
It applies to the scsi-misc-2.6 git tree.

Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

14 years agomerge by hand (qla_os.c mismerge)
James Bottomley [Sat, 18 Jun 2005 00:04:18 +0000]
merge by hand (qla_os.c mismerge)

14 years agomerge by hand (fix up qla_os.c merge error)
James Bottomley [Fri, 17 Jun 2005 23:42:23 +0000]
merge by hand (fix up qla_os.c merge error)

14 years agoLinux 2.6.12
Linus Torvalds [Fri, 17 Jun 2005 19:48:29 +0000]
Linux 2.6.12

14 years agoMerge 'for-linus' branch of rsync://rsync.kernel.org/pub/scm/linux/kernel/git/axboe...
Linus Torvalds [Fri, 17 Jun 2005 19:35:26 +0000]
Merge 'for-linus' branch of /linux/kernel/git/axboe/linux-2.6-block

14 years ago[PATCH] PCI: don't override drv->shutdown unconditionally
Christoph Hellwig [Fri, 17 Jun 2005 19:25:25 +0000]
[PATCH] PCI: don't override drv->shutdown unconditionally

There are many drivers that have been setting the generic driver
model level shutdown callback, and pci thus must not override it.

Without this patch we can have really bad data loss on various
raid controllers.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

14 years ago[SCSI] allow sleeping in ->eh_host_reset_handler()
Jeff Garzik [Sat, 28 May 2005 11:57:14 +0000]
[SCSI] allow sleeping in ->eh_host_reset_handler()

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

14 years ago[SCSI] allow sleeping in ->eh_bus_reset_handler()
Jeff Garzik [Sat, 28 May 2005 11:56:31 +0000]
[SCSI] allow sleeping in ->eh_bus_reset_handler()

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

14 years ago[SCSI] allow sleeping in ->eh_device_reset_handler()
Jeff Garzik [Sat, 28 May 2005 11:55:48 +0000]
[SCSI] allow sleeping in ->eh_device_reset_handler()

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

14 years ago[SCSI] allow sleeping in ->eh_abort_handler()
Jeff Garzik [Sat, 28 May 2005 11:54:40 +0000]
[SCSI] allow sleeping in ->eh_abort_handler()

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

14 years ago[SCSI] Remove no-op implementations of SCSI EH hooks
Jeff Garzik [Sat, 28 May 2005 11:52:51 +0000]
[SCSI] Remove no-op implementations of SCSI EH hooks

Drivers need not implement a hook that returns FAILED, and does nothing
else, since the SCSI midlayer code will do that for us.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

14 years ago[SCSI] Remove unnecessary locking around completion function calls
Jeff Garzik [Sat, 28 May 2005 11:47:39 +0000]
[SCSI] Remove unnecessary locking around completion function calls

The SCSI ->done() hook should not be called from inside a spinlock.

Drivers that do this are mostly cut-n-paste from 2.2.x-era.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

14 years ago[PATCH] timer exit cleanup
Ingo Molnar [Fri, 17 Jun 2005 09:36:36 +0000]
[PATCH] timer exit cleanup

Do all timer zapping in exit_itimers.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

14 years agoWhen cfq I/O scheduler is selected, get_request() in __make_request() calls
Kiyoshi Ueda [Fri, 17 Jun 2005 14:15:10 +0000]
When cfq I/O scheduler is selected, get_request() in __make_request() calls
__cfq_get_queue().  __cfq_get_queue() finds an existing queue (struct
cfq_queue) of the current process for the device and returns it.  If it's not
found, __cfq_get_queue() creates and returns a new one if __cfq_get_queue() is
called with __GFP_WAIT flag, or __cfq_get_queue() returns NULL (this means that
get_request() fails) if no __GFP_WAIT flag.

On the other hand, in __make_request(), get_request() is called without
__GFP_WAIT flag at the first time.  Thus, the get_request() fails when there is
no existing queue, typically when it's called for the first I/O request of the
process to the device.

Though it will be followed by get_request_wait() for general case,
__make_request() will just end the I/O with an error (EWOULDBLOCK) when the
request was for read-ahead.

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>

14 years ago[PATCH] avoid signed vs unsigned comparison in efi_range_is_wc()
Jesper Juhl [Thu, 16 Jun 2005 22:14:00 +0000]
[PATCH] avoid signed vs unsigned comparison in efi_range_is_wc()

warning when building with gcc -W :

 include/linux/efi.h: In function `efi_range_is_wc':
 include/linux/efi.h:320: warning: comparison between signed and unsigned

It looks to me like a significantly large 'len' passed in could cause the
loop to never end. Isn't it safer to make 'i' an unsigned long as well?
Like this little patch below (which of course also kills the warning) :

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Tony Luck <tony.luck@intel.com>

14 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Thu, 16 Jun 2005 20:40:39 +0000]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

14 years ago[PATCH] ARM: 2715/1: restore CPLD interrupts upon resume for Lubbock and Mainstone
Nicolas Pitre [Thu, 16 Jun 2005 20:23:56 +0000]
[PATCH] ARM: 2715/1: restore CPLD interrupts upon resume for Lubbock and Mainstone

Patch from Nicolas Pitre

Without this some devices fail to work again after a suspend event.

Signed-off-by: Nicolas Pitre
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

14 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Thu, 16 Jun 2005 20:22:23 +0000]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

14 years ago[PATCH] ARM: 2714/1: Fix the IB2 definitions for the Versatile platform
Catalin Marinas [Thu, 16 Jun 2005 17:01:12 +0000]
[PATCH] ARM: 2714/1: Fix the IB2 definitions for the Versatile platform

Patch from Catalin Marinas

The initial IB2 addresses did not depend on the IB2 base. This
patch defines them as (VERSATILE_IB2_BASE + offset).

Signed-off-by: Catalin Marinas
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

14 years ago[PATCH] ARM: 2713/1: Fix the GPIO base for Integrator/CP
Catalin Marinas [Thu, 16 Jun 2005 17:01:11 +0000]
[PATCH] ARM: 2713/1: Fix the GPIO base for Integrator/CP

Patch from Catalin Marinas

The GPIO base for Integrator/CP is different from the
Integrator/AP. This patch sets the correct value for
INTEGRATOR_GPIO_BASE.

Signed-off-by: Catalin Marinas
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

14 years ago[PATCH] ARM: 2712/1: Fix the RGB order for the Versatile CLCD
Catalin Marinas [Thu, 16 Jun 2005 17:01:11 +0000]
[PATCH] ARM: 2712/1: Fix the RGB order for the Versatile CLCD

Patch from Catalin Marinas

The current red and blue colours on the Versatile CLCD are
reversed when the 5:6:5 mode is used. The patch sets the proper
bit in the SYS_CLCD register value.

Signed-off-by: Catalin Marinas
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

14 years agoMerge 'for-linus' branch of master.kernel.org:/pub/scm/linux/kernel/git/axboe/linux...
Linus Torvalds [Thu, 16 Jun 2005 16:53:48 +0000]
Merge 'for-linus' branch of /linux/kernel/git/axboe/linux-2.6-block

14 years ago[PATCH] fix for kaweth broken by changes in the networking layer
Oliver Neukum [Thu, 16 Jun 2005 05:26:38 +0000]
[PATCH] fix for kaweth broken by changes in the networking layer

Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

14 years ago[PATCH] ALPS: fix enabling hardware tapping
Dmitry Torokhov [Thu, 16 Jun 2005 05:26:36 +0000]
[PATCH] ALPS: fix enabling hardware tapping

It looks like logic for enabling hardware tapping in ALPS driver was
inverted and we enable it only if it was already enabled by BIOS or
firmware.

I have a confirmation from one user that the patch below fixes the problem
for him and it might be beneficial if we could get it into 2.6.12.

Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

14 years ago[PATCH] Fix large core dumps with a 32-bit off_t
Daniel Jacobowitz [Thu, 16 Jun 2005 05:26:34 +0000]
[PATCH] Fix large core dumps with a 32-bit off_t

The ELF core dump code has one use of off_t when writing out segments.
Some of the segments may be passed the 2GB limit of an off_t, even on a
32-bit system, so it's important to use loff_t instead.  This fixes a
corrupted core dump in the bigcore test in GDB's testsuite.

Signed-off-by: Daniel Jacobowitz <dan@codesourcery.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

14 years ago[PATCH] sbp2 slab corruption fix
Alexandre Oliva [Thu, 16 Jun 2005 05:26:31 +0000]
[PATCH] sbp2 slab corruption fix

This fixed a problem that showed up in the Fedora development tree a few
weeks before the Fedora Core 4 release, initially as slab corruption, later
as hard crashes on boot up, when slab debugging was disabled for the
release.  More details on the history at
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=158424

The problem is caused by sbp2's use of scsi_host->hostdata[0] to hold a
scsi_id, without explicitly requesting space for it.  Since hostdata is
declared as a zero-sized array, we don't get any such space by default, so
it must be explicitly requested.  The patch below implements just that.

Signed-off-by: Alexandre Oliva <aoliva@redhat.com>
Cc: Jody McIntyre <scjody@modernduck.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

14 years ago This patch fixes q->unplug_thresh condition check in
Tejun Heo [Thu, 16 Jun 2005 10:57:31 +0000]
 This patch fixes q->unplug_thresh condition check in
__elv_add_request().  rq.count[READ] + rq.count[WRITE] can increase
more than one if another thread has allocated a request after the
current request is allocated or in_flight could have changed resulting
in larger-than-one change of nrq, thus breaking the threshold
mechanism.

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Tejun Heo <htejun@gmail.com>

14 years agoThis patch kills elevator_global_init() in elevator.c which does
Tejun Heo [Thu, 16 Jun 2005 10:56:15 +0000]
This patch kills elevator_global_init() in elevator.c which does
nothing.

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Tejun Heo <htejun@gmail.com>