linux/drivers/block/aoe/aoecmd.c

1827 lines
37 KiB
C
Raw Normal View History

/* Copyright (c) 2013 Coraid, Inc. See COPYING for GPL terms. */
/*
* aoecmd.c
* Filesystem request handling methods
*/
#include <linux/ata.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
#include <linux/slab.h>
#include <linux/hdreg.h>
#include <linux/blkdev.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/genhd.h>
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
#include <linux/moduleparam.h>
#include <linux/workqueue.h>
#include <linux/kthread.h>
[NET]: Make the device list and device lookups per namespace. This patch makes most of the generic device layer network namespace safe. This patch makes dev_base_head a network namespace variable, and then it picks up a few associated variables. The functions: dev_getbyhwaddr dev_getfirsthwbytype dev_get_by_flags dev_get_by_name __dev_get_by_name dev_get_by_index __dev_get_by_index dev_ioctl dev_ethtool dev_load wireless_process_ioctl were modified to take a network namespace argument, and deal with it. vlan_ioctl_set and brioctl_set were modified so their hooks will receive a network namespace argument. So basically anthing in the core of the network stack that was affected to by the change of dev_base was modified to handle multiple network namespaces. The rest of the network stack was simply modified to explicitly use &init_net the initial network namespace. This can be fixed when those components of the network stack are modified to handle multiple network namespaces. For now the ifindex generator is left global. Fundametally ifindex numbers are per namespace, or else we will have corner case problems with migration when we get that far. At the same time there are assumptions in the network stack that the ifindex of a network device won't change. Making the ifindex number global seems a good compromise until the network stack can cope with ifindex changes when you change namespaces, and the like. Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-09-18 02:56:21 +08:00
#include <net/net_namespace.h>
#include <asm/unaligned.h>
#include <linux/uio.h>
#include "aoe.h"
#define MAXIOC (8192) /* default meant to avoid most soft lockups */
static void ktcomplete(struct frame *, struct sk_buff *);
static int count_targets(struct aoedev *d, int *untainted);
static struct buf *nextbuf(struct aoedev *);
static int aoe_deadsecs = 60 * 3;
module_param(aoe_deadsecs, int, 0644);
MODULE_PARM_DESC(aoe_deadsecs, "After aoe_deadsecs seconds, give up and fail dev.");
static int aoe_maxout = 64;
module_param(aoe_maxout, int, 0644);
MODULE_PARM_DESC(aoe_maxout,
"Only aoe_maxout outstanding packets for every MAC on eX.Y.");
/* The number of online cpus during module initialization gives us a
* convenient heuristic cap on the parallelism used for ktio threads
* doing I/O completion. It is not important that the cap equal the
* actual number of running CPUs at any given time, but because of CPU
* hotplug, we take care to use ncpus instead of using
* num_online_cpus() after module initialization.
*/
static int ncpus;
/* mutex lock used for synchronization while thread spawning */
static DEFINE_MUTEX(ktio_spawn_lock);
static wait_queue_head_t *ktiowq;
static struct ktstate *kts;
/* io completion queue */
struct iocq_ktio {
struct list_head head;
spinlock_t lock;
};
static struct iocq_ktio *iocq;
static struct page *empty_page;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
static struct sk_buff *
new_skb(ulong len)
{
struct sk_buff *skb;
skb = alloc_skb(len + MAX_HEADER, GFP_ATOMIC);
if (skb) {
skb_reserve(skb, MAX_HEADER);
skb_reset_mac_header(skb);
skb_reset_network_header(skb);
skb->protocol = __constant_htons(ETH_P_AOE);
skb_checksum_none_assert(skb);
}
return skb;
}
static struct frame *
getframe_deferred(struct aoedev *d, u32 tag)
{
struct list_head *head, *pos, *nx;
struct frame *f;
head = &d->rexmitq;
list_for_each_safe(pos, nx, head) {
f = list_entry(pos, struct frame, head);
if (f->tag == tag) {
list_del(pos);
return f;
}
}
return NULL;
}
static struct frame *
getframe(struct aoedev *d, u32 tag)
{
struct frame *f;
struct list_head *head, *pos, *nx;
u32 n;
n = tag % NFACTIVE;
head = &d->factive[n];
list_for_each_safe(pos, nx, head) {
f = list_entry(pos, struct frame, head);
if (f->tag == tag) {
list_del(pos);
return f;
}
}
return NULL;
}
/*
* Leave the top bit clear so we have tagspace for userland.
* The bottom 16 bits are the xmit tick for rexmit/rttavg processing.
* This driver reserves tag -1 to mean "unused frame."
*/
static int
newtag(struct aoedev *d)
{
register ulong n;
n = jiffies & 0xffff;
return n |= (++d->lasttag & 0x7fff) << 16;
}
static u32
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
aoehdr_atainit(struct aoedev *d, struct aoetgt *t, struct aoe_hdr *h)
{
u32 host_tag = newtag(d);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
memcpy(h->src, t->ifp->nd->dev_addr, sizeof h->src);
memcpy(h->dst, t->addr, sizeof h->dst);
h->type = __constant_cpu_to_be16(ETH_P_AOE);
h->verfl = AOE_HVER;
h->major = cpu_to_be16(d->aoemajor);
h->minor = d->aoeminor;
h->cmd = AOECMD_ATA;
h->tag = cpu_to_be32(host_tag);
return host_tag;
}
static inline void
put_lba(struct aoe_atahdr *ah, sector_t lba)
{
ah->lba0 = lba;
ah->lba1 = lba >>= 8;
ah->lba2 = lba >>= 8;
ah->lba3 = lba >>= 8;
ah->lba4 = lba >>= 8;
ah->lba5 = lba >>= 8;
}
static struct aoeif *
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
ifrotate(struct aoetgt *t)
{
struct aoeif *ifp;
ifp = t->ifp;
ifp++;
if (ifp >= &t->ifs[NAOEIFS] || ifp->nd == NULL)
ifp = t->ifs;
if (ifp->nd == NULL)
return NULL;
return t->ifp = ifp;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
}
aoe: dynamically allocate a capped number of skbs when necessary What this Patch Does Even before this recent series of 12 patches to 2.6.22-rc4, the aoe driver was reusing a small set of skbs that were allocated once and were only used for outbound AoE commands. The network layer cannot be allowed to put_page on the data that is still associated with a bio we haven't returned to the block layer, so the aoe driver (even before the patch under discussion) is still the owner of skbs that have been handed to the network layer for transmission. We need to keep track of these skbs so that we can free them, but by tracking them, we can also easily re-use them. The new patch was a response to the behavior of certain network drivers. We cannot reuse an skb that the network driver still has in its transmit ring. Network drivers can defer transmit ring cleanup and then use the state in the skb to determine how many data segments to clean up in its transmit ring. The tg3 driver is one driver that behaves in this way. When the network driver defers cleanup of its transmit ring, the aoe driver can find itself in a situation where it would like to send an AoE command, and the AoE target is ready for more work, but the network driver still has all of the pre-allocated skbs. In that case, the new patch just calls alloc_skb, as you'd expect. We don't want to get carried away, though. We try not to do excessive allocation in the write path, so we cap the number of skbs we dynamically allocate. Probably calling it a "dynamic pool" is misleading. We were already trying to use a small fixed-size set of pre-allocated skbs before this patch, and this patch just provides a little headroom (with a ceiling, though) to accomodate network drivers that hang onto skbs, by allocating when needed. The d->skbpool_hd list of allocated skbs is necessary so that we can free them later. We didn't notice the need for this headroom until AoE targets got fast enough. Alternatives If the network layer never did a put_page on the pages in the bio's we get from the block layer, then it would be possible for us to hand skbs to the network layer and forget about them, allowing the network layer to free skbs itself (and thereby calling our own skb->destructor callback function if we needed that). In that case we could get rid of the pre-allocated skbs and also the d->skbpool_hd, instead just calling alloc_skb every time we wanted to transmit a packet. The slab allocator would effectively maintain the list of skbs. Besides a loss of CPU cache locality, the main concern with that approach the danger that it would increase the likelihood of deadlock when VM is trying to free pages by writing dirty data from the page cache through the aoe driver out to persistent storage on an AoE device. Right now we have a situation where we have pre-allocation that corresponds to how much we use, which seems ideal. Of course, there's still the separate issue of receiving the packets that tell us that a write has successfully completed on the AoE target. When memory is low and VM is using AoE to flush dirty data to free up pages, it would be perfect if there were a way for us to register a fast callback that could recognize write command completion responses. But I don't think the current problems with the receive side of the situation are a justification for exacerbating the problem on the transmit side. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:05 +08:00
static void
skb_pool_put(struct aoedev *d, struct sk_buff *skb)
{
__skb_queue_tail(&d->skbpool, skb);
aoe: dynamically allocate a capped number of skbs when necessary What this Patch Does Even before this recent series of 12 patches to 2.6.22-rc4, the aoe driver was reusing a small set of skbs that were allocated once and were only used for outbound AoE commands. The network layer cannot be allowed to put_page on the data that is still associated with a bio we haven't returned to the block layer, so the aoe driver (even before the patch under discussion) is still the owner of skbs that have been handed to the network layer for transmission. We need to keep track of these skbs so that we can free them, but by tracking them, we can also easily re-use them. The new patch was a response to the behavior of certain network drivers. We cannot reuse an skb that the network driver still has in its transmit ring. Network drivers can defer transmit ring cleanup and then use the state in the skb to determine how many data segments to clean up in its transmit ring. The tg3 driver is one driver that behaves in this way. When the network driver defers cleanup of its transmit ring, the aoe driver can find itself in a situation where it would like to send an AoE command, and the AoE target is ready for more work, but the network driver still has all of the pre-allocated skbs. In that case, the new patch just calls alloc_skb, as you'd expect. We don't want to get carried away, though. We try not to do excessive allocation in the write path, so we cap the number of skbs we dynamically allocate. Probably calling it a "dynamic pool" is misleading. We were already trying to use a small fixed-size set of pre-allocated skbs before this patch, and this patch just provides a little headroom (with a ceiling, though) to accomodate network drivers that hang onto skbs, by allocating when needed. The d->skbpool_hd list of allocated skbs is necessary so that we can free them later. We didn't notice the need for this headroom until AoE targets got fast enough. Alternatives If the network layer never did a put_page on the pages in the bio's we get from the block layer, then it would be possible for us to hand skbs to the network layer and forget about them, allowing the network layer to free skbs itself (and thereby calling our own skb->destructor callback function if we needed that). In that case we could get rid of the pre-allocated skbs and also the d->skbpool_hd, instead just calling alloc_skb every time we wanted to transmit a packet. The slab allocator would effectively maintain the list of skbs. Besides a loss of CPU cache locality, the main concern with that approach the danger that it would increase the likelihood of deadlock when VM is trying to free pages by writing dirty data from the page cache through the aoe driver out to persistent storage on an AoE device. Right now we have a situation where we have pre-allocation that corresponds to how much we use, which seems ideal. Of course, there's still the separate issue of receiving the packets that tell us that a write has successfully completed on the AoE target. When memory is low and VM is using AoE to flush dirty data to free up pages, it would be perfect if there were a way for us to register a fast callback that could recognize write command completion responses. But I don't think the current problems with the receive side of the situation are a justification for exacerbating the problem on the transmit side. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:05 +08:00
}
static struct sk_buff *
skb_pool_get(struct aoedev *d)
{
struct sk_buff *skb = skb_peek(&d->skbpool);
aoe: dynamically allocate a capped number of skbs when necessary What this Patch Does Even before this recent series of 12 patches to 2.6.22-rc4, the aoe driver was reusing a small set of skbs that were allocated once and were only used for outbound AoE commands. The network layer cannot be allowed to put_page on the data that is still associated with a bio we haven't returned to the block layer, so the aoe driver (even before the patch under discussion) is still the owner of skbs that have been handed to the network layer for transmission. We need to keep track of these skbs so that we can free them, but by tracking them, we can also easily re-use them. The new patch was a response to the behavior of certain network drivers. We cannot reuse an skb that the network driver still has in its transmit ring. Network drivers can defer transmit ring cleanup and then use the state in the skb to determine how many data segments to clean up in its transmit ring. The tg3 driver is one driver that behaves in this way. When the network driver defers cleanup of its transmit ring, the aoe driver can find itself in a situation where it would like to send an AoE command, and the AoE target is ready for more work, but the network driver still has all of the pre-allocated skbs. In that case, the new patch just calls alloc_skb, as you'd expect. We don't want to get carried away, though. We try not to do excessive allocation in the write path, so we cap the number of skbs we dynamically allocate. Probably calling it a "dynamic pool" is misleading. We were already trying to use a small fixed-size set of pre-allocated skbs before this patch, and this patch just provides a little headroom (with a ceiling, though) to accomodate network drivers that hang onto skbs, by allocating when needed. The d->skbpool_hd list of allocated skbs is necessary so that we can free them later. We didn't notice the need for this headroom until AoE targets got fast enough. Alternatives If the network layer never did a put_page on the pages in the bio's we get from the block layer, then it would be possible for us to hand skbs to the network layer and forget about them, allowing the network layer to free skbs itself (and thereby calling our own skb->destructor callback function if we needed that). In that case we could get rid of the pre-allocated skbs and also the d->skbpool_hd, instead just calling alloc_skb every time we wanted to transmit a packet. The slab allocator would effectively maintain the list of skbs. Besides a loss of CPU cache locality, the main concern with that approach the danger that it would increase the likelihood of deadlock when VM is trying to free pages by writing dirty data from the page cache through the aoe driver out to persistent storage on an AoE device. Right now we have a situation where we have pre-allocation that corresponds to how much we use, which seems ideal. Of course, there's still the separate issue of receiving the packets that tell us that a write has successfully completed on the AoE target. When memory is low and VM is using AoE to flush dirty data to free up pages, it would be perfect if there were a way for us to register a fast callback that could recognize write command completion responses. But I don't think the current problems with the receive side of the situation are a justification for exacerbating the problem on the transmit side. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:05 +08:00
if (skb && atomic_read(&skb_shinfo(skb)->dataref) == 1) {
__skb_unlink(skb, &d->skbpool);
aoe: dynamically allocate a capped number of skbs when necessary What this Patch Does Even before this recent series of 12 patches to 2.6.22-rc4, the aoe driver was reusing a small set of skbs that were allocated once and were only used for outbound AoE commands. The network layer cannot be allowed to put_page on the data that is still associated with a bio we haven't returned to the block layer, so the aoe driver (even before the patch under discussion) is still the owner of skbs that have been handed to the network layer for transmission. We need to keep track of these skbs so that we can free them, but by tracking them, we can also easily re-use them. The new patch was a response to the behavior of certain network drivers. We cannot reuse an skb that the network driver still has in its transmit ring. Network drivers can defer transmit ring cleanup and then use the state in the skb to determine how many data segments to clean up in its transmit ring. The tg3 driver is one driver that behaves in this way. When the network driver defers cleanup of its transmit ring, the aoe driver can find itself in a situation where it would like to send an AoE command, and the AoE target is ready for more work, but the network driver still has all of the pre-allocated skbs. In that case, the new patch just calls alloc_skb, as you'd expect. We don't want to get carried away, though. We try not to do excessive allocation in the write path, so we cap the number of skbs we dynamically allocate. Probably calling it a "dynamic pool" is misleading. We were already trying to use a small fixed-size set of pre-allocated skbs before this patch, and this patch just provides a little headroom (with a ceiling, though) to accomodate network drivers that hang onto skbs, by allocating when needed. The d->skbpool_hd list of allocated skbs is necessary so that we can free them later. We didn't notice the need for this headroom until AoE targets got fast enough. Alternatives If the network layer never did a put_page on the pages in the bio's we get from the block layer, then it would be possible for us to hand skbs to the network layer and forget about them, allowing the network layer to free skbs itself (and thereby calling our own skb->destructor callback function if we needed that). In that case we could get rid of the pre-allocated skbs and also the d->skbpool_hd, instead just calling alloc_skb every time we wanted to transmit a packet. The slab allocator would effectively maintain the list of skbs. Besides a loss of CPU cache locality, the main concern with that approach the danger that it would increase the likelihood of deadlock when VM is trying to free pages by writing dirty data from the page cache through the aoe driver out to persistent storage on an AoE device. Right now we have a situation where we have pre-allocation that corresponds to how much we use, which seems ideal. Of course, there's still the separate issue of receiving the packets that tell us that a write has successfully completed on the AoE target. When memory is low and VM is using AoE to flush dirty data to free up pages, it would be perfect if there were a way for us to register a fast callback that could recognize write command completion responses. But I don't think the current problems with the receive side of the situation are a justification for exacerbating the problem on the transmit side. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:05 +08:00
return skb;
}
if (skb_queue_len(&d->skbpool) < NSKBPOOLMAX &&
(skb = new_skb(ETH_ZLEN)))
aoe: dynamically allocate a capped number of skbs when necessary What this Patch Does Even before this recent series of 12 patches to 2.6.22-rc4, the aoe driver was reusing a small set of skbs that were allocated once and were only used for outbound AoE commands. The network layer cannot be allowed to put_page on the data that is still associated with a bio we haven't returned to the block layer, so the aoe driver (even before the patch under discussion) is still the owner of skbs that have been handed to the network layer for transmission. We need to keep track of these skbs so that we can free them, but by tracking them, we can also easily re-use them. The new patch was a response to the behavior of certain network drivers. We cannot reuse an skb that the network driver still has in its transmit ring. Network drivers can defer transmit ring cleanup and then use the state in the skb to determine how many data segments to clean up in its transmit ring. The tg3 driver is one driver that behaves in this way. When the network driver defers cleanup of its transmit ring, the aoe driver can find itself in a situation where it would like to send an AoE command, and the AoE target is ready for more work, but the network driver still has all of the pre-allocated skbs. In that case, the new patch just calls alloc_skb, as you'd expect. We don't want to get carried away, though. We try not to do excessive allocation in the write path, so we cap the number of skbs we dynamically allocate. Probably calling it a "dynamic pool" is misleading. We were already trying to use a small fixed-size set of pre-allocated skbs before this patch, and this patch just provides a little headroom (with a ceiling, though) to accomodate network drivers that hang onto skbs, by allocating when needed. The d->skbpool_hd list of allocated skbs is necessary so that we can free them later. We didn't notice the need for this headroom until AoE targets got fast enough. Alternatives If the network layer never did a put_page on the pages in the bio's we get from the block layer, then it would be possible for us to hand skbs to the network layer and forget about them, allowing the network layer to free skbs itself (and thereby calling our own skb->destructor callback function if we needed that). In that case we could get rid of the pre-allocated skbs and also the d->skbpool_hd, instead just calling alloc_skb every time we wanted to transmit a packet. The slab allocator would effectively maintain the list of skbs. Besides a loss of CPU cache locality, the main concern with that approach the danger that it would increase the likelihood of deadlock when VM is trying to free pages by writing dirty data from the page cache through the aoe driver out to persistent storage on an AoE device. Right now we have a situation where we have pre-allocation that corresponds to how much we use, which seems ideal. Of course, there's still the separate issue of receiving the packets that tell us that a write has successfully completed on the AoE target. When memory is low and VM is using AoE to flush dirty data to free up pages, it would be perfect if there were a way for us to register a fast callback that could recognize write command completion responses. But I don't think the current problems with the receive side of the situation are a justification for exacerbating the problem on the transmit side. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:05 +08:00
return skb;
aoe: dynamically allocate a capped number of skbs when necessary What this Patch Does Even before this recent series of 12 patches to 2.6.22-rc4, the aoe driver was reusing a small set of skbs that were allocated once and were only used for outbound AoE commands. The network layer cannot be allowed to put_page on the data that is still associated with a bio we haven't returned to the block layer, so the aoe driver (even before the patch under discussion) is still the owner of skbs that have been handed to the network layer for transmission. We need to keep track of these skbs so that we can free them, but by tracking them, we can also easily re-use them. The new patch was a response to the behavior of certain network drivers. We cannot reuse an skb that the network driver still has in its transmit ring. Network drivers can defer transmit ring cleanup and then use the state in the skb to determine how many data segments to clean up in its transmit ring. The tg3 driver is one driver that behaves in this way. When the network driver defers cleanup of its transmit ring, the aoe driver can find itself in a situation where it would like to send an AoE command, and the AoE target is ready for more work, but the network driver still has all of the pre-allocated skbs. In that case, the new patch just calls alloc_skb, as you'd expect. We don't want to get carried away, though. We try not to do excessive allocation in the write path, so we cap the number of skbs we dynamically allocate. Probably calling it a "dynamic pool" is misleading. We were already trying to use a small fixed-size set of pre-allocated skbs before this patch, and this patch just provides a little headroom (with a ceiling, though) to accomodate network drivers that hang onto skbs, by allocating when needed. The d->skbpool_hd list of allocated skbs is necessary so that we can free them later. We didn't notice the need for this headroom until AoE targets got fast enough. Alternatives If the network layer never did a put_page on the pages in the bio's we get from the block layer, then it would be possible for us to hand skbs to the network layer and forget about them, allowing the network layer to free skbs itself (and thereby calling our own skb->destructor callback function if we needed that). In that case we could get rid of the pre-allocated skbs and also the d->skbpool_hd, instead just calling alloc_skb every time we wanted to transmit a packet. The slab allocator would effectively maintain the list of skbs. Besides a loss of CPU cache locality, the main concern with that approach the danger that it would increase the likelihood of deadlock when VM is trying to free pages by writing dirty data from the page cache through the aoe driver out to persistent storage on an AoE device. Right now we have a situation where we have pre-allocation that corresponds to how much we use, which seems ideal. Of course, there's still the separate issue of receiving the packets that tell us that a write has successfully completed on the AoE target. When memory is low and VM is using AoE to flush dirty data to free up pages, it would be perfect if there were a way for us to register a fast callback that could recognize write command completion responses. But I don't think the current problems with the receive side of the situation are a justification for exacerbating the problem on the transmit side. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:05 +08:00
return NULL;
}
void
aoe_freetframe(struct frame *f)
{
struct aoetgt *t;
t = f->t;
f->buf = NULL;
memset(&f->iter, 0, sizeof(f->iter));
f->r_skb = NULL;
f->flags = 0;
list_add(&f->head, &t->ffree);
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
static struct frame *
newtframe(struct aoedev *d, struct aoetgt *t)
{
struct frame *f;
aoe: dynamically allocate a capped number of skbs when necessary What this Patch Does Even before this recent series of 12 patches to 2.6.22-rc4, the aoe driver was reusing a small set of skbs that were allocated once and were only used for outbound AoE commands. The network layer cannot be allowed to put_page on the data that is still associated with a bio we haven't returned to the block layer, so the aoe driver (even before the patch under discussion) is still the owner of skbs that have been handed to the network layer for transmission. We need to keep track of these skbs so that we can free them, but by tracking them, we can also easily re-use them. The new patch was a response to the behavior of certain network drivers. We cannot reuse an skb that the network driver still has in its transmit ring. Network drivers can defer transmit ring cleanup and then use the state in the skb to determine how many data segments to clean up in its transmit ring. The tg3 driver is one driver that behaves in this way. When the network driver defers cleanup of its transmit ring, the aoe driver can find itself in a situation where it would like to send an AoE command, and the AoE target is ready for more work, but the network driver still has all of the pre-allocated skbs. In that case, the new patch just calls alloc_skb, as you'd expect. We don't want to get carried away, though. We try not to do excessive allocation in the write path, so we cap the number of skbs we dynamically allocate. Probably calling it a "dynamic pool" is misleading. We were already trying to use a small fixed-size set of pre-allocated skbs before this patch, and this patch just provides a little headroom (with a ceiling, though) to accomodate network drivers that hang onto skbs, by allocating when needed. The d->skbpool_hd list of allocated skbs is necessary so that we can free them later. We didn't notice the need for this headroom until AoE targets got fast enough. Alternatives If the network layer never did a put_page on the pages in the bio's we get from the block layer, then it would be possible for us to hand skbs to the network layer and forget about them, allowing the network layer to free skbs itself (and thereby calling our own skb->destructor callback function if we needed that). In that case we could get rid of the pre-allocated skbs and also the d->skbpool_hd, instead just calling alloc_skb every time we wanted to transmit a packet. The slab allocator would effectively maintain the list of skbs. Besides a loss of CPU cache locality, the main concern with that approach the danger that it would increase the likelihood of deadlock when VM is trying to free pages by writing dirty data from the page cache through the aoe driver out to persistent storage on an AoE device. Right now we have a situation where we have pre-allocation that corresponds to how much we use, which seems ideal. Of course, there's still the separate issue of receiving the packets that tell us that a write has successfully completed on the AoE target. When memory is low and VM is using AoE to flush dirty data to free up pages, it would be perfect if there were a way for us to register a fast callback that could recognize write command completion responses. But I don't think the current problems with the receive side of the situation are a justification for exacerbating the problem on the transmit side. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:05 +08:00
struct sk_buff *skb;
struct list_head *pos;
if (list_empty(&t->ffree)) {
if (t->falloc >= NSKBPOOLMAX*2)
return NULL;
f = kcalloc(1, sizeof(*f), GFP_ATOMIC);
if (f == NULL)
return NULL;
t->falloc++;
f->t = t;
} else {
pos = t->ffree.next;
list_del(pos);
f = list_entry(pos, struct frame, head);
}
skb = f->skb;
if (skb == NULL) {
f->skb = skb = new_skb(ETH_ZLEN);
if (!skb) {
bail: aoe_freetframe(f);
return NULL;
}
}
if (atomic_read(&skb_shinfo(skb)->dataref) != 1) {
skb = skb_pool_get(d);
if (skb == NULL)
goto bail;
skb_pool_put(d, f->skb);
f->skb = skb;
}
skb->truesize -= skb->data_len;
skb_shinfo(skb)->nr_frags = skb->data_len = 0;
skb_trim(skb, 0);
return f;
}
static struct frame *
newframe(struct aoedev *d)
{
struct frame *f;
struct aoetgt *t, **tt;
int totout = 0;
int use_tainted;
int has_untainted;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
if (!d->targets || !d->targets[0]) {
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
printk(KERN_ERR "aoe: NULL TARGETS!\n");
return NULL;
}
tt = d->tgt; /* last used target */
for (use_tainted = 0, has_untainted = 0;;) {
tt++;
if (tt >= &d->targets[d->ntargets] || !*tt)
tt = d->targets;
t = *tt;
if (!t->taint) {
has_untainted = 1;
totout += t->nout;
}
if (t->nout < t->maxout
&& (use_tainted || !t->taint)
&& t->ifp->nd) {
f = newtframe(d, t);
if (f) {
ifrotate(t);
d->tgt = tt;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
return f;
}
}
if (tt == d->tgt) { /* we've looped and found nada */
if (!use_tainted && !has_untainted)
use_tainted = 1;
else
break;
}
}
if (totout == 0) {
d->kicked++;
d->flags |= DEVFL_KICKME;
aoe: dynamically allocate a capped number of skbs when necessary What this Patch Does Even before this recent series of 12 patches to 2.6.22-rc4, the aoe driver was reusing a small set of skbs that were allocated once and were only used for outbound AoE commands. The network layer cannot be allowed to put_page on the data that is still associated with a bio we haven't returned to the block layer, so the aoe driver (even before the patch under discussion) is still the owner of skbs that have been handed to the network layer for transmission. We need to keep track of these skbs so that we can free them, but by tracking them, we can also easily re-use them. The new patch was a response to the behavior of certain network drivers. We cannot reuse an skb that the network driver still has in its transmit ring. Network drivers can defer transmit ring cleanup and then use the state in the skb to determine how many data segments to clean up in its transmit ring. The tg3 driver is one driver that behaves in this way. When the network driver defers cleanup of its transmit ring, the aoe driver can find itself in a situation where it would like to send an AoE command, and the AoE target is ready for more work, but the network driver still has all of the pre-allocated skbs. In that case, the new patch just calls alloc_skb, as you'd expect. We don't want to get carried away, though. We try not to do excessive allocation in the write path, so we cap the number of skbs we dynamically allocate. Probably calling it a "dynamic pool" is misleading. We were already trying to use a small fixed-size set of pre-allocated skbs before this patch, and this patch just provides a little headroom (with a ceiling, though) to accomodate network drivers that hang onto skbs, by allocating when needed. The d->skbpool_hd list of allocated skbs is necessary so that we can free them later. We didn't notice the need for this headroom until AoE targets got fast enough. Alternatives If the network layer never did a put_page on the pages in the bio's we get from the block layer, then it would be possible for us to hand skbs to the network layer and forget about them, allowing the network layer to free skbs itself (and thereby calling our own skb->destructor callback function if we needed that). In that case we could get rid of the pre-allocated skbs and also the d->skbpool_hd, instead just calling alloc_skb every time we wanted to transmit a packet. The slab allocator would effectively maintain the list of skbs. Besides a loss of CPU cache locality, the main concern with that approach the danger that it would increase the likelihood of deadlock when VM is trying to free pages by writing dirty data from the page cache through the aoe driver out to persistent storage on an AoE device. Right now we have a situation where we have pre-allocation that corresponds to how much we use, which seems ideal. Of course, there's still the separate issue of receiving the packets that tell us that a write has successfully completed on the AoE target. When memory is low and VM is using AoE to flush dirty data to free up pages, it would be perfect if there were a way for us to register a fast callback that could recognize write command completion responses. But I don't think the current problems with the receive side of the situation are a justification for exacerbating the problem on the transmit side. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:05 +08:00
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
return NULL;
}
static void
skb_fillup(struct sk_buff *skb, struct bio *bio, struct bvec_iter iter)
{
int frag = 0;
struct bio_vec bv;
__bio_for_each_segment(bv, bio, iter, iter)
skb_fill_page_desc(skb, frag++, bv.bv_page,
bv.bv_offset, bv.bv_len);
}
static void
fhash(struct frame *f)
{
struct aoedev *d = f->t->d;
u32 n;
n = f->tag % NFACTIVE;
list_add_tail(&f->head, &d->factive[n]);
}
static void
ata_rw_frameinit(struct frame *f)
{
struct aoetgt *t;
struct aoe_hdr *h;
struct aoe_atahdr *ah;
struct sk_buff *skb;
char writebit, extbit;
skb = f->skb;
h = (struct aoe_hdr *) skb_mac_header(skb);
ah = (struct aoe_atahdr *) (h + 1);
skb_put(skb, sizeof(*h) + sizeof(*ah));
memset(h, 0, skb->len);
writebit = 0x10;
extbit = 0x4;
t = f->t;
f->tag = aoehdr_atainit(t->d, t, h);
fhash(f);
t->nout++;
f->waited = 0;
f->waited_total = 0;
/* set up ata header */
ah->scnt = f->iter.bi_size >> 9;
put_lba(ah, f->iter.bi_sector);
if (t->d->flags & DEVFL_EXT) {
ah->aflags |= AOEAFL_EXT;
} else {
extbit = 0;
ah->lba3 &= 0x0f;
ah->lba3 |= 0xe0; /* LBA bit + obsolete 0xa0 */
}
if (f->buf && bio_data_dir(f->buf->bio) == WRITE) {
skb_fillup(skb, f->buf->bio, f->iter);
ah->aflags |= AOEAFL_WRITE;
skb->len += f->iter.bi_size;
skb->data_len = f->iter.bi_size;
skb->truesize += f->iter.bi_size;
t->wpkts++;
} else {
t->rpkts++;
writebit = 0;
}
ah->cmdstat = ATA_CMD_PIO_READ | writebit | extbit;
skb->dev = t->ifp->nd;
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
static int
aoecmd_ata_rw(struct aoedev *d)
{
struct frame *f;
struct buf *buf;
struct sk_buff *skb;
struct sk_buff_head queue;
buf = nextbuf(d);
if (buf == NULL)
return 0;
f = newframe(d);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
if (f == NULL)
return 0;
/* initialize the headers & frame */
f->buf = buf;
f->iter = buf->iter;
f->iter.bi_size = min_t(unsigned long,
d->maxbcnt ?: DEFAULTBCNT,
f->iter.bi_size);
bio_advance_iter(buf->bio, &buf->iter, f->iter.bi_size);
if (!buf->iter.bi_size)
d->ip.buf = NULL;
/* mark all tracking fields and load out */
buf->nframesout += 1;
ata_rw_frameinit(f);
skb = skb_clone(f->skb, GFP_ATOMIC);
if (skb) {
do_gettimeofday(&f->sent);
f->sent_jiffs = (u32) jiffies;
__skb_queue_head_init(&queue);
__skb_queue_tail(&queue, skb);
aoenet_xmit(&queue);
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
return 1;
}
/* some callers cannot sleep, and they can call this function,
* transmitting the packets later, when interrupts are on
*/
static void
aoecmd_cfg_pkts(ushort aoemajor, unsigned char aoeminor, struct sk_buff_head *queue)
{
struct aoe_hdr *h;
struct aoe_cfghdr *ch;
struct sk_buff *skb;
struct net_device *ifp;
rcu_read_lock();
for_each_netdev_rcu(&init_net, ifp) {
dev_hold(ifp);
if (!is_aoe_netif(ifp))
goto cont;
skb = new_skb(sizeof *h + sizeof *ch);
if (skb == NULL) {
printk(KERN_INFO "aoe: skb alloc failure\n");
goto cont;
}
skb_put(skb, sizeof *h + sizeof *ch);
skb->dev = ifp;
__skb_queue_tail(queue, skb);
h = (struct aoe_hdr *) skb_mac_header(skb);
memset(h, 0, sizeof *h + sizeof *ch);
memset(h->dst, 0xff, sizeof h->dst);
memcpy(h->src, ifp->dev_addr, sizeof h->src);
h->type = __constant_cpu_to_be16(ETH_P_AOE);
h->verfl = AOE_HVER;
h->major = cpu_to_be16(aoemajor);
h->minor = aoeminor;
h->cmd = AOECMD_CFG;
cont:
dev_put(ifp);
}
rcu_read_unlock();
}
static void
resend(struct aoedev *d, struct frame *f)
{
struct sk_buff *skb;
struct sk_buff_head queue;
struct aoe_hdr *h;
struct aoetgt *t;
char buf[128];
u32 n;
t = f->t;
n = newtag(d);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
skb = f->skb;
if (ifrotate(t) == NULL) {
/* probably can't happen, but set it up to fail anyway */
pr_info("aoe: resend: no interfaces to rotate to.\n");
ktcomplete(f, NULL);
return;
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
h = (struct aoe_hdr *) skb_mac_header(skb);
if (!(f->flags & FFL_PROBE)) {
snprintf(buf, sizeof(buf),
"%15s e%ld.%d oldtag=%08x@%08lx newtag=%08x s=%pm d=%pm nout=%d\n",
"retransmit", d->aoemajor, d->aoeminor,
f->tag, jiffies, n,
h->src, h->dst, t->nout);
aoechr_error(buf);
}
f->tag = n;
fhash(f);
h->tag = cpu_to_be32(n);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
memcpy(h->dst, t->addr, sizeof h->dst);
memcpy(h->src, t->ifp->nd->dev_addr, sizeof h->src);
skb->dev = t->ifp->nd;
skb = skb_clone(skb, GFP_ATOMIC);
if (skb == NULL)
return;
do_gettimeofday(&f->sent);
f->sent_jiffs = (u32) jiffies;
__skb_queue_head_init(&queue);
__skb_queue_tail(&queue, skb);
aoenet_xmit(&queue);
}
static int
tsince_hr(struct frame *f)
{
struct timeval now;
int n;
do_gettimeofday(&now);
n = now.tv_usec - f->sent.tv_usec;
n += (now.tv_sec - f->sent.tv_sec) * USEC_PER_SEC;
if (n < 0)
n = -n;
/* For relatively long periods, use jiffies to avoid
* discrepancies caused by updates to the system time.
*
* On system with HZ of 1000, 32-bits is over 49 days
* worth of jiffies, or over 71 minutes worth of usecs.
*
* Jiffies overflow is handled by subtraction of unsigned ints:
* (gdb) print (unsigned) 2 - (unsigned) 0xfffffffe
* $3 = 4
* (gdb)
*/
if (n > USEC_PER_SEC / 4) {
n = ((u32) jiffies) - f->sent_jiffs;
n *= USEC_PER_SEC / HZ;
}
return n;
}
static int
tsince(u32 tag)
{
int n;
n = jiffies & 0xffff;
n -= tag & 0xffff;
if (n < 0)
n += 1<<16;
return jiffies_to_usecs(n + 1);
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
static struct aoeif *
getif(struct aoetgt *t, struct net_device *nd)
{
struct aoeif *p, *e;
p = t->ifs;
e = p + NAOEIFS;
for (; p < e; p++)
if (p->nd == nd)
return p;
return NULL;
}
static void
ejectif(struct aoetgt *t, struct aoeif *ifp)
{
struct aoeif *e;
struct net_device *nd;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
ulong n;
nd = ifp->nd;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
e = t->ifs + NAOEIFS - 1;
n = (e - ifp) * sizeof *ifp;
memmove(ifp, ifp+1, n);
e->nd = NULL;
dev_put(nd);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
}
static struct frame *
reassign_frame(struct frame *f)
{
struct frame *nf;
struct sk_buff *skb;
nf = newframe(f->t->d);
if (!nf)
return NULL;
if (nf->t == f->t) {
aoe_freetframe(nf);
return NULL;
}
skb = nf->skb;
nf->skb = f->skb;
nf->buf = f->buf;
nf->iter = f->iter;
nf->waited = 0;
nf->waited_total = f->waited_total;
nf->sent = f->sent;
nf->sent_jiffs = f->sent_jiffs;
f->skb = skb;
return nf;
}
static void
probe(struct aoetgt *t)
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
{
struct aoedev *d;
struct frame *f;
struct sk_buff *skb;
struct sk_buff_head queue;
size_t n, m;
int frag;
d = t->d;
f = newtframe(d, t);
if (!f) {
pr_err("%s %pm for e%ld.%d: %s\n",
"aoe: cannot probe remote address",
t->addr,
(long) d->aoemajor, d->aoeminor,
"no frame available");
return;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
}
f->flags |= FFL_PROBE;
ifrotate(t);
f->iter.bi_size = t->d->maxbcnt ? t->d->maxbcnt : DEFAULTBCNT;
ata_rw_frameinit(f);
skb = f->skb;
for (frag = 0, n = f->iter.bi_size; n > 0; ++frag, n -= m) {
if (n < PAGE_SIZE)
m = n;
else
m = PAGE_SIZE;
skb_fill_page_desc(skb, frag, empty_page, 0, m);
}
skb->len += f->iter.bi_size;
skb->data_len = f->iter.bi_size;
skb->truesize += f->iter.bi_size;
skb = skb_clone(f->skb, GFP_ATOMIC);
if (skb) {
do_gettimeofday(&f->sent);
f->sent_jiffs = (u32) jiffies;
__skb_queue_head_init(&queue);
__skb_queue_tail(&queue, skb);
aoenet_xmit(&queue);
}
}
static long
rto(struct aoedev *d)
{
long t;
t = 2 * d->rttavg >> RTTSCALE;
t += 8 * d->rttdev >> RTTDSCALE;
if (t == 0)
t = 1;
return t;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
}
static void
rexmit_deferred(struct aoedev *d)
{
struct aoetgt *t;
struct frame *f;
struct frame *nf;
struct list_head *pos, *nx, *head;
int since;
int untainted;
count_targets(d, &untainted);
head = &d->rexmitq;
list_for_each_safe(pos, nx, head) {
f = list_entry(pos, struct frame, head);
t = f->t;
if (t->taint) {
if (!(f->flags & FFL_PROBE)) {
nf = reassign_frame(f);
if (nf) {
if (t->nout_probes == 0
&& untainted > 0) {
probe(t);
t->nout_probes++;
}
list_replace(&f->head, &nf->head);
pos = &nf->head;
aoe_freetframe(f);
f = nf;
t = f->t;
}
} else if (untainted < 1) {
/* don't probe w/o other untainted aoetgts */
goto stop_probe;
} else if (tsince_hr(f) < t->taint * rto(d)) {
/* reprobe slowly when taint is high */
continue;
}
} else if (f->flags & FFL_PROBE) {
stop_probe: /* don't probe untainted aoetgts */
list_del(pos);
aoe_freetframe(f);
/* leaving d->kicked, because this is routine */
f->t->d->flags |= DEVFL_KICKME;
continue;
}
if (t->nout >= t->maxout)
continue;
list_del(pos);
t->nout++;
if (f->flags & FFL_PROBE)
t->nout_probes++;
since = tsince_hr(f);
f->waited += since;
f->waited_total += since;
resend(d, f);
}
}
/* An aoetgt accumulates demerits quickly, and successful
* probing redeems the aoetgt slowly.
*/
static void
scorn(struct aoetgt *t)
{
int n;
n = t->taint++;
t->taint += t->taint * 2;
if (n > t->taint)
t->taint = n;
if (t->taint > MAX_TAINT)
t->taint = MAX_TAINT;
}
static int
count_targets(struct aoedev *d, int *untainted)
{
int i, good;
for (i = good = 0; i < d->ntargets && d->targets[i]; ++i)
if (d->targets[i]->taint == 0)
good++;
if (untainted)
*untainted = good;
return i;
}
static void
rexmit_timer(ulong vp)
{
struct aoedev *d;
struct aoetgt *t;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
struct aoeif *ifp;
struct frame *f;
struct list_head *head, *pos, *nx;
LIST_HEAD(flist);
register long timeout;
ulong flags, n;
int i;
int utgts; /* number of aoetgt descriptors (not slots) */
int since;
d = (struct aoedev *) vp;
spin_lock_irqsave(&d->lock, flags);
/* timeout based on observed timings and variations */
timeout = rto(d);
utgts = count_targets(d, NULL);
if (d->flags & DEVFL_TKILL) {
spin_unlock_irqrestore(&d->lock, flags);
return;
}
/* collect all frames to rexmit into flist */
for (i = 0; i < NFACTIVE; i++) {
head = &d->factive[i];
list_for_each_safe(pos, nx, head) {
f = list_entry(pos, struct frame, head);
if (tsince_hr(f) < timeout)
break; /* end of expired frames */
/* move to flist for later processing */
list_move_tail(pos, &flist);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
}
}
/* process expired frames */
while (!list_empty(&flist)) {
pos = flist.next;
f = list_entry(pos, struct frame, head);
since = tsince_hr(f);
n = f->waited_total + since;
n /= USEC_PER_SEC;
if (aoe_deadsecs
&& n > aoe_deadsecs
&& !(f->flags & FFL_PROBE)) {
/* Waited too long. Device failure.
* Hang all frames on first hash bucket for downdev
* to clean up.
*/
list_splice(&flist, &d->factive[0]);
aoedev_downdev(d);
goto out;
}
t = f->t;
n = f->waited + since;
n /= USEC_PER_SEC;
if (aoe_deadsecs && utgts > 0
&& (n > aoe_deadsecs / utgts || n > HARD_SCORN_SECS))
scorn(t); /* avoid this target */
if (t->maxout != 1) {
t->ssthresh = t->maxout / 2;
t->maxout = 1;
}
if (f->flags & FFL_PROBE) {
t->nout_probes--;
} else {
ifp = getif(t, f->skb->dev);
if (ifp && ++ifp->lost > (t->nframes << 1)
&& (ifp != t->ifs || t->ifs[1].nd)) {
ejectif(t, ifp);
ifp = NULL;
}
}
list_move_tail(pos, &d->rexmitq);
t->nout--;
}
rexmit_deferred(d);
out:
if ((d->flags & DEVFL_KICKME) && d->blkq) {
d->flags &= ~DEVFL_KICKME;
d->blkq->request_fn(d->blkq);
}
d->timer.expires = jiffies + TIMERTICK;
add_timer(&d->timer);
spin_unlock_irqrestore(&d->lock, flags);
}
static unsigned long
rqbiocnt(struct request *r)
{
struct bio *bio;
unsigned long n = 0;
__rq_for_each_bio(bio, r)
n++;
return n;
}
/* This can be removed if we are certain that no users of the block
* layer will ever use zero-count pages in bios. Otherwise we have to
* protect against the put_page sometimes done by the network layer.
*
* See http://oss.sgi.com/archives/xfs/2007-01/msg00594.html for
* discussion.
*
* We cannot use get_page in the workaround, because it insists on a
* positive page count as a precondition. So we use _count directly.
*/
static void
bio_pageinc(struct bio *bio)
{
block: Convert bio_for_each_segment() to bvec_iter More prep work for immutable biovecs - with immutable bvecs drivers won't be able to use the biovec directly, they'll need to use helpers that take into account bio->bi_iter.bi_bvec_done. This updates callers for the new usage without changing the implementation yet. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Paul Clements <Paul.Clements@steeleye.com> Cc: Jim Paris <jim@jtan.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Nagalakshmi Nandigama <Nagalakshmi.Nandigama@lsi.com> Cc: Sreekanth Reddy <Sreekanth.Reddy@lsi.com> Cc: support@lsi.com Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Tejun Heo <tj@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Matthew Wilcox <matthew.r.wilcox@intel.com> Cc: Keith Busch <keith.busch@intel.com> Cc: Stephen Hemminger <shemminger@vyatta.com> Cc: Quoc-Son Anh <quoc-sonx.anh@intel.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Seth Jennings <sjenning@linux.vnet.ibm.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Jan Kara <jack@suse.cz> Cc: linux-m68k@lists.linux-m68k.org Cc: linuxppc-dev@lists.ozlabs.org Cc: drbd-user@lists.linbit.com Cc: nbd-general@lists.sourceforge.net Cc: cbe-oss-dev@lists.ozlabs.org Cc: xen-devel@lists.xensource.com Cc: virtualization@lists.linux-foundation.org Cc: linux-raid@vger.kernel.org Cc: linux-s390@vger.kernel.org Cc: DL-MPTFusionLinux@lsi.com Cc: linux-scsi@vger.kernel.org Cc: devel@driverdev.osuosl.org Cc: linux-fsdevel@vger.kernel.org Cc: cluster-devel@redhat.com Cc: linux-mm@kvack.org Acked-by: Geoff Levand <geoff@infradead.org>
2013-11-24 09:19:00 +08:00
struct bio_vec bv;
struct page *page;
block: Convert bio_for_each_segment() to bvec_iter More prep work for immutable biovecs - with immutable bvecs drivers won't be able to use the biovec directly, they'll need to use helpers that take into account bio->bi_iter.bi_bvec_done. This updates callers for the new usage without changing the implementation yet. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Paul Clements <Paul.Clements@steeleye.com> Cc: Jim Paris <jim@jtan.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Nagalakshmi Nandigama <Nagalakshmi.Nandigama@lsi.com> Cc: Sreekanth Reddy <Sreekanth.Reddy@lsi.com> Cc: support@lsi.com Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Tejun Heo <tj@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Matthew Wilcox <matthew.r.wilcox@intel.com> Cc: Keith Busch <keith.busch@intel.com> Cc: Stephen Hemminger <shemminger@vyatta.com> Cc: Quoc-Son Anh <quoc-sonx.anh@intel.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Seth Jennings <sjenning@linux.vnet.ibm.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Jan Kara <jack@suse.cz> Cc: linux-m68k@lists.linux-m68k.org Cc: linuxppc-dev@lists.ozlabs.org Cc: drbd-user@lists.linbit.com Cc: nbd-general@lists.sourceforge.net Cc: cbe-oss-dev@lists.ozlabs.org Cc: xen-devel@lists.xensource.com Cc: virtualization@lists.linux-foundation.org Cc: linux-raid@vger.kernel.org Cc: linux-s390@vger.kernel.org Cc: DL-MPTFusionLinux@lsi.com Cc: linux-scsi@vger.kernel.org Cc: devel@driverdev.osuosl.org Cc: linux-fsdevel@vger.kernel.org Cc: cluster-devel@redhat.com Cc: linux-mm@kvack.org Acked-by: Geoff Levand <geoff@infradead.org>
2013-11-24 09:19:00 +08:00
struct bvec_iter iter;
block: Convert bio_for_each_segment() to bvec_iter More prep work for immutable biovecs - with immutable bvecs drivers won't be able to use the biovec directly, they'll need to use helpers that take into account bio->bi_iter.bi_bvec_done. This updates callers for the new usage without changing the implementation yet. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Paul Clements <Paul.Clements@steeleye.com> Cc: Jim Paris <jim@jtan.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Nagalakshmi Nandigama <Nagalakshmi.Nandigama@lsi.com> Cc: Sreekanth Reddy <Sreekanth.Reddy@lsi.com> Cc: support@lsi.com Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Tejun Heo <tj@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Matthew Wilcox <matthew.r.wilcox@intel.com> Cc: Keith Busch <keith.busch@intel.com> Cc: Stephen Hemminger <shemminger@vyatta.com> Cc: Quoc-Son Anh <quoc-sonx.anh@intel.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Seth Jennings <sjenning@linux.vnet.ibm.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Jan Kara <jack@suse.cz> Cc: linux-m68k@lists.linux-m68k.org Cc: linuxppc-dev@lists.ozlabs.org Cc: drbd-user@lists.linbit.com Cc: nbd-general@lists.sourceforge.net Cc: cbe-oss-dev@lists.ozlabs.org Cc: xen-devel@lists.xensource.com Cc: virtualization@lists.linux-foundation.org Cc: linux-raid@vger.kernel.org Cc: linux-s390@vger.kernel.org Cc: DL-MPTFusionLinux@lsi.com Cc: linux-scsi@vger.kernel.org Cc: devel@driverdev.osuosl.org Cc: linux-fsdevel@vger.kernel.org Cc: cluster-devel@redhat.com Cc: linux-mm@kvack.org Acked-by: Geoff Levand <geoff@infradead.org>
2013-11-24 09:19:00 +08:00
bio_for_each_segment(bv, bio, iter) {
/* Non-zero page count for non-head members of
* compound pages is no longer allowed by the kernel.
*/
mm: close PageTail race Commit bf6bddf1924e ("mm: introduce compaction and migration for ballooned pages") introduces page_count(page) into memory compaction which dereferences page->first_page if PageTail(page). This results in a very rare NULL pointer dereference on the aforementioned page_count(page). Indeed, anything that does compound_head(), including page_count() is susceptible to racing with prep_compound_page() and seeing a NULL or dangling page->first_page pointer. This patch uses Andrea's implementation of compound_trans_head() that deals with such a race and makes it the default compound_head() implementation. This includes a read memory barrier that ensures that if PageTail(head) is true that we return a head page that is neither NULL nor dangling. The patch then adds a store memory barrier to prep_compound_page() to ensure page->first_page is set. This is the safest way to ensure we see the head page that we are expecting, PageTail(page) is already in the unlikely() path and the memory barriers are unfortunately required. Hugetlbfs is the exception, we don't enforce a store memory barrier during init since no race is possible. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Holger Kiehl <Holger.Kiehl@dwd.de> Cc: Christoph Lameter <cl@linux.com> Cc: Rafael Aquini <aquini@redhat.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Michal Hocko <mhocko@suse.cz> Cc: Mel Gorman <mgorman@suse.de> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-03-04 07:38:18 +08:00
page = compound_head(bv.bv_page);
atomic_inc(&page->_count);
}
}
static void
bio_pagedec(struct bio *bio)
{
struct page *page;
block: Convert bio_for_each_segment() to bvec_iter More prep work for immutable biovecs - with immutable bvecs drivers won't be able to use the biovec directly, they'll need to use helpers that take into account bio->bi_iter.bi_bvec_done. This updates callers for the new usage without changing the implementation yet. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Paul Clements <Paul.Clements@steeleye.com> Cc: Jim Paris <jim@jtan.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Nagalakshmi Nandigama <Nagalakshmi.Nandigama@lsi.com> Cc: Sreekanth Reddy <Sreekanth.Reddy@lsi.com> Cc: support@lsi.com Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Tejun Heo <tj@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Matthew Wilcox <matthew.r.wilcox@intel.com> Cc: Keith Busch <keith.busch@intel.com> Cc: Stephen Hemminger <shemminger@vyatta.com> Cc: Quoc-Son Anh <quoc-sonx.anh@intel.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Seth Jennings <sjenning@linux.vnet.ibm.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Jan Kara <jack@suse.cz> Cc: linux-m68k@lists.linux-m68k.org Cc: linuxppc-dev@lists.ozlabs.org Cc: drbd-user@lists.linbit.com Cc: nbd-general@lists.sourceforge.net Cc: cbe-oss-dev@lists.ozlabs.org Cc: xen-devel@lists.xensource.com Cc: virtualization@lists.linux-foundation.org Cc: linux-raid@vger.kernel.org Cc: linux-s390@vger.kernel.org Cc: DL-MPTFusionLinux@lsi.com Cc: linux-scsi@vger.kernel.org Cc: devel@driverdev.osuosl.org Cc: linux-fsdevel@vger.kernel.org Cc: cluster-devel@redhat.com Cc: linux-mm@kvack.org Acked-by: Geoff Levand <geoff@infradead.org>
2013-11-24 09:19:00 +08:00
struct bio_vec bv;
struct bvec_iter iter;
block: Convert bio_for_each_segment() to bvec_iter More prep work for immutable biovecs - with immutable bvecs drivers won't be able to use the biovec directly, they'll need to use helpers that take into account bio->bi_iter.bi_bvec_done. This updates callers for the new usage without changing the implementation yet. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Paul Clements <Paul.Clements@steeleye.com> Cc: Jim Paris <jim@jtan.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Nagalakshmi Nandigama <Nagalakshmi.Nandigama@lsi.com> Cc: Sreekanth Reddy <Sreekanth.Reddy@lsi.com> Cc: support@lsi.com Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Tejun Heo <tj@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Matthew Wilcox <matthew.r.wilcox@intel.com> Cc: Keith Busch <keith.busch@intel.com> Cc: Stephen Hemminger <shemminger@vyatta.com> Cc: Quoc-Son Anh <quoc-sonx.anh@intel.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Seth Jennings <sjenning@linux.vnet.ibm.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Jan Kara <jack@suse.cz> Cc: linux-m68k@lists.linux-m68k.org Cc: linuxppc-dev@lists.ozlabs.org Cc: drbd-user@lists.linbit.com Cc: nbd-general@lists.sourceforge.net Cc: cbe-oss-dev@lists.ozlabs.org Cc: xen-devel@lists.xensource.com Cc: virtualization@lists.linux-foundation.org Cc: linux-raid@vger.kernel.org Cc: linux-s390@vger.kernel.org Cc: DL-MPTFusionLinux@lsi.com Cc: linux-scsi@vger.kernel.org Cc: devel@driverdev.osuosl.org Cc: linux-fsdevel@vger.kernel.org Cc: cluster-devel@redhat.com Cc: linux-mm@kvack.org Acked-by: Geoff Levand <geoff@infradead.org>
2013-11-24 09:19:00 +08:00
bio_for_each_segment(bv, bio, iter) {
mm: close PageTail race Commit bf6bddf1924e ("mm: introduce compaction and migration for ballooned pages") introduces page_count(page) into memory compaction which dereferences page->first_page if PageTail(page). This results in a very rare NULL pointer dereference on the aforementioned page_count(page). Indeed, anything that does compound_head(), including page_count() is susceptible to racing with prep_compound_page() and seeing a NULL or dangling page->first_page pointer. This patch uses Andrea's implementation of compound_trans_head() that deals with such a race and makes it the default compound_head() implementation. This includes a read memory barrier that ensures that if PageTail(head) is true that we return a head page that is neither NULL nor dangling. The patch then adds a store memory barrier to prep_compound_page() to ensure page->first_page is set. This is the safest way to ensure we see the head page that we are expecting, PageTail(page) is already in the unlikely() path and the memory barriers are unfortunately required. Hugetlbfs is the exception, we don't enforce a store memory barrier during init since no race is possible. Signed-off-by: David Rientjes <rientjes@google.com> Cc: Holger Kiehl <Holger.Kiehl@dwd.de> Cc: Christoph Lameter <cl@linux.com> Cc: Rafael Aquini <aquini@redhat.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Michal Hocko <mhocko@suse.cz> Cc: Mel Gorman <mgorman@suse.de> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Rik van Riel <riel@redhat.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-03-04 07:38:18 +08:00
page = compound_head(bv.bv_page);
atomic_dec(&page->_count);
}
}
static void
bufinit(struct buf *buf, struct request *rq, struct bio *bio)
{
memset(buf, 0, sizeof(*buf));
buf->rq = rq;
buf->bio = bio;
buf->iter = bio->bi_iter;
bio_pageinc(bio);
}
static struct buf *
nextbuf(struct aoedev *d)
{
struct request *rq;
struct request_queue *q;
struct buf *buf;
struct bio *bio;
q = d->blkq;
if (q == NULL)
return NULL; /* initializing */
if (d->ip.buf)
return d->ip.buf;
rq = d->ip.rq;
if (rq == NULL) {
rq = blk_peek_request(q);
if (rq == NULL)
return NULL;
blk_start_request(rq);
d->ip.rq = rq;
d->ip.nxbio = rq->bio;
rq->special = (void *) rqbiocnt(rq);
}
buf = mempool_alloc(d->bufpool, GFP_ATOMIC);
if (buf == NULL) {
pr_err("aoe: nextbuf: unable to mempool_alloc!\n");
return NULL;
}
bio = d->ip.nxbio;
bufinit(buf, rq, bio);
bio = bio->bi_next;
d->ip.nxbio = bio;
if (bio == NULL)
d->ip.rq = NULL;
return d->ip.buf = buf;
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
/* enters with d->lock held */
void
aoecmd_work(struct aoedev *d)
{
rexmit_deferred(d);
while (aoecmd_ata_rw(d))
;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
}
/* this function performs work that has been deferred until sleeping is OK
*/
void
aoecmd_sleepwork(struct work_struct *work)
{
struct aoedev *d = container_of(work, struct aoedev, work);
struct block_device *bd;
u64 ssize;
if (d->flags & DEVFL_GDALLOC)
aoeblk_gdalloc(d);
if (d->flags & DEVFL_NEWSIZE) {
ssize = get_capacity(d->gd);
bd = bdget_disk(d->gd, 0);
if (bd) {
mutex_lock(&bd->bd_inode->i_mutex);
i_size_write(bd->bd_inode, (loff_t)ssize<<9);
mutex_unlock(&bd->bd_inode->i_mutex);
bdput(bd);
}
spin_lock_irq(&d->lock);
d->flags |= DEVFL_UP;
d->flags &= ~DEVFL_NEWSIZE;
spin_unlock_irq(&d->lock);
}
}
static void
ata_ident_fixstring(u16 *id, int ns)
{
u16 s;
while (ns-- > 0) {
s = *id;
*id++ = s >> 8 | s << 8;
}
}
static void
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
ataid_complete(struct aoedev *d, struct aoetgt *t, unsigned char *id)
{
u64 ssize;
u16 n;
/* word 83: command set supported */
n = get_unaligned_le16(&id[83 << 1]);
/* word 86: command set/feature enabled */
n |= get_unaligned_le16(&id[86 << 1]);
if (n & (1<<10)) { /* bit 10: LBA 48 */
d->flags |= DEVFL_EXT;
/* word 100: number lba48 sectors */
ssize = get_unaligned_le64(&id[100 << 1]);
/* set as in ide-disk.c:init_idedisk_capacity */
d->geo.cylinders = ssize;
d->geo.cylinders /= (255 * 63);
d->geo.heads = 255;
d->geo.sectors = 63;
} else {
d->flags &= ~DEVFL_EXT;
/* number lba28 sectors */
ssize = get_unaligned_le32(&id[60 << 1]);
/* NOTE: obsolete in ATA 6 */
d->geo.cylinders = get_unaligned_le16(&id[54 << 1]);
d->geo.heads = get_unaligned_le16(&id[55 << 1]);
d->geo.sectors = get_unaligned_le16(&id[56 << 1]);
}
ata_ident_fixstring((u16 *) &id[10<<1], 10); /* serial */
ata_ident_fixstring((u16 *) &id[23<<1], 4); /* firmware */
ata_ident_fixstring((u16 *) &id[27<<1], 20); /* model */
memcpy(d->ident, id, sizeof(d->ident));
if (d->ssize != ssize)
printk(KERN_INFO
"aoe: %pm e%ld.%d v%04x has %llu sectors\n",
t->addr,
d->aoemajor, d->aoeminor,
d->fw_ver, (long long)ssize);
d->ssize = ssize;
d->geo.start = 0;
if (d->flags & (DEVFL_GDALLOC|DEVFL_NEWSIZE))
return;
if (d->gd != NULL) {
set_capacity(d->gd, ssize);
d->flags |= DEVFL_NEWSIZE;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
} else
d->flags |= DEVFL_GDALLOC;
schedule_work(&d->work);
}
static void
calc_rttavg(struct aoedev *d, struct aoetgt *t, int rtt)
{
register long n;
n = rtt;
/* cf. Congestion Avoidance and Control, Jacobson & Karels, 1988 */
n -= d->rttavg >> RTTSCALE;
d->rttavg += n;
if (n < 0)
n = -n;
n -= d->rttdev >> RTTDSCALE;
d->rttdev += n;
if (!t || t->maxout >= t->nframes)
return;
if (t->maxout < t->ssthresh)
t->maxout += 1;
else if (t->nout == t->maxout && t->next_cwnd-- == 0) {
t->maxout += 1;
t->next_cwnd = t->maxout;
}
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
static struct aoetgt *
gettgt(struct aoedev *d, char *addr)
{
struct aoetgt **t, **e;
t = d->targets;
e = t + d->ntargets;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
for (; t < e && *t; t++)
if (memcmp((*t)->addr, addr, sizeof((*t)->addr)) == 0)
return *t;
return NULL;
}
static void
bvcpy(struct sk_buff *skb, struct bio *bio, struct bvec_iter iter, long cnt)
{
int soff = 0;
struct bio_vec bv;
iter.bi_size = cnt;
__bio_for_each_segment(bv, bio, iter, iter) {
char *p = page_address(bv.bv_page) + bv.bv_offset;
skb_copy_bits(skb, soff, p, bv.bv_len);
soff += bv.bv_len;
}
}
void
aoe_end_request(struct aoedev *d, struct request *rq, int fastfail)
{
struct bio *bio;
int bok;
struct request_queue *q;
q = d->blkq;
if (rq == d->ip.rq)
d->ip.rq = NULL;
do {
bio = rq->bio;
bok = !fastfail && !bio->bi_error;
block: Abstract out bvec iterator Immutable biovecs are going to require an explicit iterator. To implement immutable bvecs, a later patch is going to add a bi_bvec_done member to this struct; for now, this patch effectively just renames things. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Matthew Wilcox <willy@linux.intel.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Alasdair Kergon <agk@redhat.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: dm-devel@redhat.com Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Boaz Harrosh <bharrosh@panasas.com> Cc: Benny Halevy <bhalevy@tonian.com> Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: "Nicholas A. Bellinger" <nab@linux-iscsi.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Chris Mason <chris.mason@fusionio.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jaegeuk Kim <jaegeuk.kim@samsung.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@kernel.org> Cc: Joern Engel <joern@logfs.org> Cc: Prasad Joshi <prasadjoshi.linux@gmail.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: KONISHI Ryusuke <konishi.ryusuke@lab.ntt.co.jp> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Ben Myers <bpm@sgi.com> Cc: xfs@oss.sgi.com Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Len Brown <len.brown@intel.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Wei Yongjun <yongjun_wei@trendmicro.com.cn> Cc: "Roger Pau Monné" <roger.pau@citrix.com> Cc: Jan Beulich <jbeulich@suse.com> Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Cc: Ian Campbell <Ian.Campbell@citrix.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Jiang Liu <jiang.liu@huawei.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchand@redhat.com> Cc: Joe Perches <joe@perches.com> Cc: Peng Tao <tao.peng@emc.com> Cc: Andy Adamson <andros@netapp.com> Cc: fanchaoting <fanchaoting@cn.fujitsu.com> Cc: Jie Liu <jeff.liu@oracle.com> Cc: Sunil Mushran <sunil.mushran@gmail.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Namjae Jeon <namjae.jeon@samsung.com> Cc: Pankaj Kumar <pankaj.km@samsung.com> Cc: Dan Magenheimer <dan.magenheimer@oracle.com> Cc: Mel Gorman <mgorman@suse.de>6
2013-10-12 06:44:27 +08:00
} while (__blk_end_request(rq, bok ? 0 : -EIO, bio->bi_iter.bi_size));
/* cf. http://lkml.org/lkml/2006/10/31/28 */
if (!fastfail)
__blk_run_queue(q);
}
static void
aoe_end_buf(struct aoedev *d, struct buf *buf)
{
struct request *rq;
unsigned long n;
if (buf == d->ip.buf)
d->ip.buf = NULL;
rq = buf->rq;
bio_pagedec(buf->bio);
mempool_free(buf, d->bufpool);
n = (unsigned long) rq->special;
rq->special = (void *) --n;
if (n == 0)
aoe_end_request(d, rq, 0);
}
static void
ktiocomplete(struct frame *f)
{
struct aoe_hdr *hin, *hout;
struct aoe_atahdr *ahin, *ahout;
struct buf *buf;
struct sk_buff *skb;
struct aoetgt *t;
struct aoeif *ifp;
struct aoedev *d;
long n;
int untainted;
if (f == NULL)
return;
t = f->t;
d = t->d;
skb = f->r_skb;
buf = f->buf;
if (f->flags & FFL_PROBE)
goto out;
if (!skb) /* just fail the buf. */
goto noskb;
hout = (struct aoe_hdr *) skb_mac_header(f->skb);
ahout = (struct aoe_atahdr *) (hout+1);
hin = (struct aoe_hdr *) skb->data;
skb_pull(skb, sizeof(*hin));
ahin = (struct aoe_atahdr *) skb->data;
skb_pull(skb, sizeof(*ahin));
if (ahin->cmdstat & 0xa9) { /* these bits cleared on success */
pr_err("aoe: ata error cmd=%2.2Xh stat=%2.2Xh from e%ld.%d\n",
ahout->cmdstat, ahin->cmdstat,
d->aoemajor, d->aoeminor);
noskb: if (buf)
buf->bio->bi_error = -EIO;
goto out;
}
n = ahout->scnt << 9;
switch (ahout->cmdstat) {
case ATA_CMD_PIO_READ:
case ATA_CMD_PIO_READ_EXT:
if (skb->len < n) {
pr_err("%s e%ld.%d. skb->len=%d need=%ld\n",
"aoe: runt data size in read from",
(long) d->aoemajor, d->aoeminor,
skb->len, n);
buf->bio->bi_error = -EIO;
break;
}
if (n > f->iter.bi_size) {
pr_err_ratelimited("%s e%ld.%d. bytes=%ld need=%u\n",
"aoe: too-large data size in read from",
(long) d->aoemajor, d->aoeminor,
n, f->iter.bi_size);
buf->bio->bi_error = -EIO;
break;
}
bvcpy(skb, f->buf->bio, f->iter, n);
case ATA_CMD_PIO_WRITE:
case ATA_CMD_PIO_WRITE_EXT:
spin_lock_irq(&d->lock);
ifp = getif(t, skb->dev);
if (ifp)
ifp->lost = 0;
spin_unlock_irq(&d->lock);
break;
case ATA_CMD_ID_ATA:
if (skb->len < 512) {
pr_info("%s e%ld.%d. skb->len=%d need=512\n",
"aoe: runt data size in ataid from",
(long) d->aoemajor, d->aoeminor,
skb->len);
break;
}
if (skb_linearize(skb))
break;
spin_lock_irq(&d->lock);
ataid_complete(d, t, skb->data);
spin_unlock_irq(&d->lock);
break;
default:
pr_info("aoe: unrecognized ata command %2.2Xh for %d.%d\n",
ahout->cmdstat,
be16_to_cpu(get_unaligned(&hin->major)),
hin->minor);
}
out:
spin_lock_irq(&d->lock);
if (t->taint > 0
&& --t->taint > 0
&& t->nout_probes == 0) {
count_targets(d, &untainted);
if (untainted > 0) {
probe(t);
t->nout_probes++;
}
}
aoe_freetframe(f);
if (buf && --buf->nframesout == 0 && buf->iter.bi_size == 0)
aoe_end_buf(d, buf);
spin_unlock_irq(&d->lock);
aoedev_put(d);
dev_kfree_skb(skb);
}
/* Enters with iocq.lock held.
* Returns true iff responses needing processing remain.
*/
static int
ktio(int id)
{
struct frame *f;
struct list_head *pos;
int i;
int actual_id;
for (i = 0; ; ++i) {
if (i == MAXIOC)
return 1;
if (list_empty(&iocq[id].head))
return 0;
pos = iocq[id].head.next;
list_del(pos);
f = list_entry(pos, struct frame, head);
spin_unlock_irq(&iocq[id].lock);
ktiocomplete(f);
/* Figure out if extra threads are required. */
actual_id = f->t->d->aoeminor % ncpus;
if (!kts[actual_id].active) {
BUG_ON(id != 0);
mutex_lock(&ktio_spawn_lock);
if (!kts[actual_id].active
&& aoe_ktstart(&kts[actual_id]) == 0)
kts[actual_id].active = 1;
mutex_unlock(&ktio_spawn_lock);
}
spin_lock_irq(&iocq[id].lock);
}
}
static int
kthread(void *vp)
{
struct ktstate *k;
DECLARE_WAITQUEUE(wait, current);
int more;
k = vp;
current->flags |= PF_NOFREEZE;
set_user_nice(current, -10);
complete(&k->rendez); /* tell spawner we're running */
do {
spin_lock_irq(k->lock);
more = k->fn(k->id);
if (!more) {
add_wait_queue(k->waitq, &wait);
__set_current_state(TASK_INTERRUPTIBLE);
}
spin_unlock_irq(k->lock);
if (!more) {
schedule();
remove_wait_queue(k->waitq, &wait);
} else
cond_resched();
} while (!kthread_should_stop());
complete(&k->rendez); /* tell spawner we're stopping */
return 0;
}
void
aoe_ktstop(struct ktstate *k)
{
kthread_stop(k->task);
wait_for_completion(&k->rendez);
}
int
aoe_ktstart(struct ktstate *k)
{
struct task_struct *task;
init_completion(&k->rendez);
task = kthread_run(kthread, k, "%s", k->name);
if (task == NULL || IS_ERR(task))
return -ENOMEM;
k->task = task;
wait_for_completion(&k->rendez); /* allow kthread to start */
init_completion(&k->rendez); /* for waiting for exit later */
return 0;
}
/* pass it off to kthreads for processing */
static void
ktcomplete(struct frame *f, struct sk_buff *skb)
{
int id;
ulong flags;
f->r_skb = skb;
id = f->t->d->aoeminor % ncpus;
spin_lock_irqsave(&iocq[id].lock, flags);
if (!kts[id].active) {
spin_unlock_irqrestore(&iocq[id].lock, flags);
/* The thread with id has not been spawned yet,
* so delegate the work to the main thread and
* try spawning a new thread.
*/
id = 0;
spin_lock_irqsave(&iocq[id].lock, flags);
}
list_add_tail(&f->head, &iocq[id].head);
spin_unlock_irqrestore(&iocq[id].lock, flags);
wake_up(&ktiowq[id]);
}
struct sk_buff *
aoecmd_ata_rsp(struct sk_buff *skb)
{
struct aoedev *d;
struct aoe_hdr *h;
struct frame *f;
u32 n;
ulong flags;
char ebuf[128];
u16 aoemajor;
h = (struct aoe_hdr *) skb->data;
aoemajor = be16_to_cpu(get_unaligned(&h->major));
d = aoedev_by_aoeaddr(aoemajor, h->minor, 0);
if (d == NULL) {
snprintf(ebuf, sizeof ebuf, "aoecmd_ata_rsp: ata response "
"for unknown device %d.%d\n",
aoemajor, h->minor);
aoechr_error(ebuf);
return skb;
}
spin_lock_irqsave(&d->lock, flags);
n = be32_to_cpu(get_unaligned(&h->tag));
f = getframe(d, n);
if (f) {
calc_rttavg(d, f->t, tsince_hr(f));
f->t->nout--;
if (f->flags & FFL_PROBE)
f->t->nout_probes--;
} else {
f = getframe_deferred(d, n);
if (f) {
calc_rttavg(d, NULL, tsince_hr(f));
} else {
calc_rttavg(d, NULL, tsince(n));
spin_unlock_irqrestore(&d->lock, flags);
aoedev_put(d);
snprintf(ebuf, sizeof(ebuf),
"%15s e%d.%d tag=%08x@%08lx s=%pm d=%pm\n",
"unexpected rsp",
get_unaligned_be16(&h->major),
h->minor,
get_unaligned_be32(&h->tag),
jiffies,
h->src,
h->dst);
aoechr_error(ebuf);
return skb;
}
}
aoecmd_work(d);
spin_unlock_irqrestore(&d->lock, flags);
ktcomplete(f, skb);
/*
* Note here that we do not perform an aoedev_put, as we are
* leaving this reference for the ktio to release.
*/
return NULL;
}
void
aoecmd_cfg(ushort aoemajor, unsigned char aoeminor)
{
struct sk_buff_head queue;
__skb_queue_head_init(&queue);
aoecmd_cfg_pkts(aoemajor, aoeminor, &queue);
aoenet_xmit(&queue);
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
struct sk_buff *
aoecmd_ata_id(struct aoedev *d)
{
struct aoe_hdr *h;
struct aoe_atahdr *ah;
struct frame *f;
struct sk_buff *skb;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
struct aoetgt *t;
f = newframe(d);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
if (f == NULL)
return NULL;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
t = *d->tgt;
/* initialize the headers & frame */
skb = f->skb;
h = (struct aoe_hdr *) skb_mac_header(skb);
ah = (struct aoe_atahdr *) (h+1);
skb_put(skb, sizeof *h + sizeof *ah);
memset(h, 0, skb->len);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
f->tag = aoehdr_atainit(d, t, h);
fhash(f);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
t->nout++;
f->waited = 0;
f->waited_total = 0;
/* set up ata header */
ah->scnt = 1;
ah->cmdstat = ATA_CMD_ID_ATA;
ah->lba3 = 0xa0;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
skb->dev = t->ifp->nd;
d->rttavg = RTTAVG_INIT;
d->rttdev = RTTDEV_INIT;
d->timer.function = rexmit_timer;
skb = skb_clone(skb, GFP_ATOMIC);
if (skb) {
do_gettimeofday(&f->sent);
f->sent_jiffs = (u32) jiffies;
}
return skb;
}
static struct aoetgt **
grow_targets(struct aoedev *d)
{
ulong oldn, newn;
struct aoetgt **tt;
oldn = d->ntargets;
newn = oldn * 2;
tt = kcalloc(newn, sizeof(*d->targets), GFP_ATOMIC);
if (!tt)
return NULL;
memmove(tt, d->targets, sizeof(*d->targets) * oldn);
d->tgt = tt + (d->tgt - d->targets);
kfree(d->targets);
d->targets = tt;
d->ntargets = newn;
return &d->targets[oldn];
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
static struct aoetgt *
addtgt(struct aoedev *d, char *addr, ulong nframes)
{
struct aoetgt *t, **tt, **te;
tt = d->targets;
te = tt + d->ntargets;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
for (; tt < te && *tt; tt++)
;
if (tt == te) {
tt = grow_targets(d);
if (!tt)
goto nomem;
}
t = kzalloc(sizeof(*t), GFP_ATOMIC);
if (!t)
goto nomem;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
t->nframes = nframes;
t->d = d;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
memcpy(t->addr, addr, sizeof t->addr);
t->ifp = t->ifs;
aoecmd_wreset(t);
t->maxout = t->nframes / 2;
INIT_LIST_HEAD(&t->ffree);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
return *tt = t;
nomem:
pr_info("aoe: cannot allocate memory to add target\n");
return NULL;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
}
static void
setdbcnt(struct aoedev *d)
{
struct aoetgt **t, **e;
int bcnt = 0;
t = d->targets;
e = t + d->ntargets;
for (; t < e && *t; t++)
if (bcnt == 0 || bcnt > (*t)->minbcnt)
bcnt = (*t)->minbcnt;
if (bcnt != d->maxbcnt) {
d->maxbcnt = bcnt;
pr_info("aoe: e%ld.%d: setting %d byte data frames\n",
d->aoemajor, d->aoeminor, bcnt);
}
}
static void
setifbcnt(struct aoetgt *t, struct net_device *nd, int bcnt)
{
struct aoedev *d;
struct aoeif *p, *e;
int minbcnt;
d = t->d;
minbcnt = bcnt;
p = t->ifs;
e = p + NAOEIFS;
for (; p < e; p++) {
if (p->nd == NULL)
break; /* end of the valid interfaces */
if (p->nd == nd) {
p->bcnt = bcnt; /* we're updating */
nd = NULL;
} else if (minbcnt > p->bcnt)
minbcnt = p->bcnt; /* find the min interface */
}
if (nd) {
if (p == e) {
pr_err("aoe: device setifbcnt failure; too many interfaces.\n");
return;
}
dev_hold(nd);
p->nd = nd;
p->bcnt = bcnt;
}
t->minbcnt = minbcnt;
setdbcnt(d);
}
void
aoecmd_cfg_rsp(struct sk_buff *skb)
{
struct aoedev *d;
struct aoe_hdr *h;
struct aoe_cfghdr *ch;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
struct aoetgt *t;
ulong flags, aoemajor;
struct sk_buff *sl;
struct sk_buff_head queue;
u16 n;
sl = NULL;
h = (struct aoe_hdr *) skb_mac_header(skb);
ch = (struct aoe_cfghdr *) (h+1);
/*
* Enough people have their dip switches set backwards to
* warrant a loud message for this special case.
*/
aoemajor = get_unaligned_be16(&h->major);
if (aoemajor == 0xfff) {
printk(KERN_ERR "aoe: Warning: shelf address is all ones. "
"Check shelf dip switches.\n");
return;
}
if (aoemajor == 0xffff) {
pr_info("aoe: e%ld.%d: broadcast shelf number invalid\n",
aoemajor, (int) h->minor);
return;
}
if (h->minor == 0xff) {
pr_info("aoe: e%ld.%d: broadcast slot number invalid\n",
aoemajor, (int) h->minor);
return;
}
n = be16_to_cpu(ch->bufcnt);
if (n > aoe_maxout) /* keep it reasonable */
n = aoe_maxout;
d = aoedev_by_aoeaddr(aoemajor, h->minor, 1);
if (d == NULL) {
pr_info("aoe: device allocation failure\n");
return;
}
spin_lock_irqsave(&d->lock, flags);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
t = gettgt(d, h->src);
if (t) {
t->nframes = n;
if (n < t->maxout)
aoecmd_wreset(t);
} else {
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
t = addtgt(d, h->src, n);
if (!t)
goto bail;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
}
n = skb->dev->mtu;
n -= sizeof(struct aoe_hdr) + sizeof(struct aoe_atahdr);
n /= 512;
if (n > ch->scnt)
n = ch->scnt;
n = n ? n * 512 : DEFAULTBCNT;
setifbcnt(t, skb->dev, n);
/* don't change users' perspective */
if (d->nopen == 0) {
d->fw_ver = be16_to_cpu(ch->fwver);
sl = aoecmd_ata_id(d);
}
bail:
spin_unlock_irqrestore(&d->lock, flags);
aoedev_put(d);
if (sl) {
__skb_queue_head_init(&queue);
__skb_queue_tail(&queue, sl);
aoenet_xmit(&queue);
}
}
void
aoecmd_wreset(struct aoetgt *t)
{
t->maxout = 1;
t->ssthresh = t->nframes / 2;
t->next_cwnd = t->nframes;
}
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
void
aoecmd_cleanslate(struct aoedev *d)
{
struct aoetgt **t, **te;
d->rttavg = RTTAVG_INIT;
d->rttdev = RTTDEV_INIT;
d->maxbcnt = 0;
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
t = d->targets;
te = t + d->ntargets;
for (; t < te && *t; t++)
aoecmd_wreset(*t);
aoe: handle multiple network paths to AoE device A remote AoE device is something can process ATA commands and is identified by an AoE shelf number and an AoE slot number. Such a device might have more than one network interface, and it might be reachable by more than one local network interface. This patch tracks the available network paths available to each AoE device, allowing them to be used more efficiently. Andrew Morton asked about the call to msleep_interruptible in the revalidate function. Yes, if a signal is pending, then msleep_interruptible will not return 0. That means we will not loop but will call aoenet_xmit with a NULL skb, which is a noop. If the system is too low on memory or the aoe driver is too low on frames, then the user can hit control-C to interrupt the attempt to do a revalidate. I have added a comment to the code summarizing that. Andrew Morton asked whether the allocation performed inside addtgt could use a more relaxed allocation like GFP_KERNEL, but addtgt is called when the aoedev lock has been locked with spin_lock_irqsave. It would be nice to allocate the memory under fewer restrictions, but targets are only added when the device is being discovered, and if the target can't be added right now, we can try again in a minute when then next AoE config query broadcast goes out. Andrew Morton pointed out that the "too many targets" message could be printed for failing GFP_ATOMIC allocations. The last patch in this series makes the messages more specific. Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-02-08 20:20:00 +08:00
}
void
aoe_failbuf(struct aoedev *d, struct buf *buf)
{
if (buf == NULL)
return;
buf->iter.bi_size = 0;
buf->bio->bi_error = -EIO;
if (buf->nframesout == 0)
aoe_end_buf(d, buf);
}
void
aoe_flush_iocq(void)
{
int i;
for (i = 0; i < ncpus; i++) {
if (kts[i].active)
aoe_flush_iocq_by_index(i);
}
}
void
aoe_flush_iocq_by_index(int id)
{
struct frame *f;
struct aoedev *d;
LIST_HEAD(flist);
struct list_head *pos;
struct sk_buff *skb;
ulong flags;
spin_lock_irqsave(&iocq[id].lock, flags);
list_splice_init(&iocq[id].head, &flist);
spin_unlock_irqrestore(&iocq[id].lock, flags);
while (!list_empty(&flist)) {
pos = flist.next;
list_del(pos);
f = list_entry(pos, struct frame, head);
d = f->t->d;
skb = f->r_skb;
spin_lock_irqsave(&d->lock, flags);
if (f->buf) {
f->buf->nframesout--;
aoe_failbuf(d, f->buf);
}
aoe_freetframe(f);
spin_unlock_irqrestore(&d->lock, flags);
dev_kfree_skb(skb);
aoedev_put(d);
}
}
int __init
aoecmd_init(void)
{
void *p;
int i;
int ret;
/* get_zeroed_page returns page with ref count 1 */
p = (void *) get_zeroed_page(GFP_KERNEL | __GFP_REPEAT);
if (!p)
return -ENOMEM;
empty_page = virt_to_page(p);
ncpus = num_online_cpus();
iocq = kcalloc(ncpus, sizeof(struct iocq_ktio), GFP_KERNEL);
if (!iocq)
return -ENOMEM;
kts = kcalloc(ncpus, sizeof(struct ktstate), GFP_KERNEL);
if (!kts) {
ret = -ENOMEM;
goto kts_fail;
}
ktiowq = kcalloc(ncpus, sizeof(wait_queue_head_t), GFP_KERNEL);
if (!ktiowq) {
ret = -ENOMEM;
goto ktiowq_fail;
}
mutex_init(&ktio_spawn_lock);
for (i = 0; i < ncpus; i++) {
INIT_LIST_HEAD(&iocq[i].head);
spin_lock_init(&iocq[i].lock);
init_waitqueue_head(&ktiowq[i]);
snprintf(kts[i].name, sizeof(kts[i].name), "aoe_ktio%d", i);
kts[i].fn = ktio;
kts[i].waitq = &ktiowq[i];
kts[i].lock = &iocq[i].lock;
kts[i].id = i;
kts[i].active = 0;
}
kts[0].active = 1;
if (aoe_ktstart(&kts[0])) {
ret = -ENOMEM;
goto ktstart_fail;
}
return 0;
ktstart_fail:
kfree(ktiowq);
ktiowq_fail:
kfree(kts);
kts_fail:
kfree(iocq);
return ret;
}
void
aoecmd_exit(void)
{
int i;
for (i = 0; i < ncpus; i++)
if (kts[i].active)
aoe_ktstop(&kts[i]);
aoe_flush_iocq();
/* Free up the iocq and thread speicific configuration
* allocated during startup.
*/
kfree(iocq);
kfree(kts);
kfree(ktiowq);
free_page((unsigned long) page_address(empty_page));
empty_page = NULL;
}