Re: [PATCH] Marvell Libertas 8388 802.11b/g USB driver (v2)
On Wed, Jan 17, 2007 at 06:19:07PM -0500, Jeff Garzik wrote: > Christoph Hellwig wrote: > >On Wed, Jan 17, 2007 at 01:00:47PM -0500, Dan Williams wrote: > >>allows the 8388 to continue routing other laptops' packets over the mesh > >>*while the host CPU is asleep*. > > > >We're not going to put a lot of junk into the kernel just because the OLPC > >folks decide to do odd powermanagment schemes. > > We're not going to ignore useful power management schemes just because > they don't fit neatly into a pre-existing category. > > I think the request to determine how all this maps into MLME is fair, > though. Definitely. Also, I wonder if there was any attempt to evaluate how the ieee80211 (or d80211) code might be extended in order to elimnate the need for some of the libertas wlan_* files? John -- John W. Linville [EMAIL PROTECTED] - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: watchdog timeout panic in e1000 driver
Hi, > My patch may seem like a huge change, but in essence the change is > pretty simple. > > In my patch, the interrupt handler code will check whether the interrupted > code is holding the swfw semaphore. If it is held, the watchdog function > is deferred until swfw semaphore is released. > The modification is for the interrupted code which is holding the > semaphore, and the interrupt handler, so they are both directly related > to this problem. > > I will try to add some comments to my code to make it more readable. I have rebased this patch for 2.6.20-rc5 and added some comments to this patch to make it more readable. Does this patch have problems? Or do you have any other ideas? -- Kenzo Iwami ([EMAIL PROTECTED]) Signed-off-by: Kenzo Iwami <[EMAIL PROTECTED]> diff -urpN linux-2.6.20-rc5_org/drivers/net/e1000/e1000_hw.c linux-2.6.20-rc5_fix/drivers/net/e1000/e1000_hw.c --- linux-2.6.20-rc5_org/drivers/net/e1000/e1000_hw.c 2007-01-13 03:54:26.0 +0900 +++ linux-2.6.20-rc5_fix/drivers/net/e1000/e1000_hw.c 2007-01-18 18:04:04.0 +0900 @@ -35,6 +35,7 @@ static int32_t e1000_swfw_sync_acquire(struct e1000_hw *hw, uint16_t mask); static void e1000_swfw_sync_release(struct e1000_hw *hw, uint16_t mask); +static void e1000_check_watchdog_deferred(struct e1000_hw *hw); static int32_t e1000_read_kmrn_reg(struct e1000_hw *hw, uint32_t reg_addr, uint16_t *data); static int32_t e1000_write_kmrn_reg(struct e1000_hw *hw, uint32_t reg_addr, uint16_t data); static int32_t e1000_get_software_semaphore(struct e1000_hw *hw); @@ -3396,6 +3397,29 @@ e1000_shift_in_mdi_bits(struct e1000_hw return data; } +static void +e1000_check_watchdog_deferred(struct e1000_hw *hw) +{ +/* If watchdog interrupt was deferred while this process was holding + * the swfw semaphore, process the watchdog now. + */ +if (atomic_xchg(&hw->watchdog_deferred, 0)) { +retry: +e1000_do_watchdog(hw); +} +atomic_dec(&hw->swfw_sem_count); + +/* Check watchdog_deferred once more just in case watchdog interrupt + * occurred between atomic_xchg and atomic_dec. + * By nature of the watchdog interrupt, we shouldn't go through this + * more than once. + */ +if (atomic_xchg(&hw->watchdog_deferred, 0)) { +atomic_inc(&hw->swfw_sem_count); +goto retry; +} +} + static int32_t e1000_swfw_sync_acquire(struct e1000_hw *hw, uint16_t mask) { @@ -3413,8 +3437,11 @@ e1000_swfw_sync_acquire(struct e1000_hw return e1000_get_hw_eeprom_semaphore(hw); while (timeout) { -if (e1000_get_hw_eeprom_semaphore(hw)) +atomic_inc(&hw->swfw_sem_count); +if (e1000_get_hw_eeprom_semaphore(hw)) { +e1000_check_watchdog_deferred(hw); return -E1000_ERR_SWFW_SYNC; +} swfw_sync = E1000_READ_REG(hw, SW_FW_SYNC); if (!(swfw_sync & (fwmask | swmask))) { @@ -3424,6 +3451,7 @@ e1000_swfw_sync_acquire(struct e1000_hw /* firmware currently using resource (fwmask) */ /* or other software thread currently using resource (swmask) */ e1000_put_hw_eeprom_semaphore(hw); +e1000_check_watchdog_deferred(hw); mdelay(5); timeout--; } @@ -3437,6 +3465,8 @@ e1000_swfw_sync_acquire(struct e1000_hw E1000_WRITE_REG(hw, SW_FW_SYNC, swfw_sync); e1000_put_hw_eeprom_semaphore(hw); +e1000_check_watchdog_deferred(hw); + return E1000_SUCCESS; } @@ -3458,6 +3488,7 @@ e1000_swfw_sync_release(struct e1000_hw return; } +atomic_inc(&hw->swfw_sem_count); /* if (e1000_get_hw_eeprom_semaphore(hw)) *return -E1000_ERR_SWFW_SYNC; */ while (e1000_get_hw_eeprom_semaphore(hw) != E1000_SUCCESS); @@ -3468,6 +3499,7 @@ e1000_swfw_sync_release(struct e1000_hw E1000_WRITE_REG(hw, SW_FW_SYNC, swfw_sync); e1000_put_hw_eeprom_semaphore(hw); +e1000_check_watchdog_deferred(hw); } /* diff -urpN linux-2.6.20-rc5_org/drivers/net/e1000/e1000_hw.h linux-2.6.20-rc5_fix/drivers/net/e1000/e1000_hw.h --- linux-2.6.20-rc5_org/drivers/net/e1000/e1000_hw.h 2007-01-13 03:54:26.0 +0900 +++ linux-2.6.20-rc5_fix/drivers/net/e1000/e1000_hw.h 2007-01-18 18:04:42.0 +0900 @@ -306,6 +306,7 @@ typedef enum { #define E1000_BYTE_SWAP_WORD(_value) _value) & 0x00ff) << 8) | \ (((_value) & 0xff00) >> 8)) +extern void e1000_do_watchdog(struct e1000_hw *hw); /* Function prototypes */ /* Initialization */ int32_t e1000_reset_hw(struct e1000_hw *hw); @@ -1465,6 +1466,8 @@ struct e1000_hw { boolean_t has_manc2h; boolean_t rx_needs_kicking; boolean_t has_smbus; + atomic_tswfw_sem_count;/* >0 if swfw_sem held on ESB2 */ + atomic_t
Re: [PATCH 1/12] L2 network namespace (v3): current network namespace operations
On Wednesday 17 January 2007 23:16, Eric W. Biederman wrote: > Dmitry Mishin <[EMAIL PROTECTED]> writes: > > > Added functions and macros required to operate with network namespaces. > > They are required in order to switch network namespace for incoming packets > > and > > to not extend current network interface by additional network namespace > > argue. > > Is exec_net only used in interrupt context? I tried to do so. > Or how do you ensure a sleeping function does not get called and the > kernel process comes back on another cpu? Seems that I forgot to remove it's usage at least in one place - in clone_net_ns(). If you caught more, please, let me know. -- Thanks, Dmitry. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Can someone please try...
On Wed, 2007-01-17 at 10:52 +0100, Michael Buesch wrote: > Doesn't happen for me. I have no idea what's happening. > Care to debug it? > But it's weird that _killing_ the supplicant calls add_interface. > I'd expect it to call remove_interface. I'm sorry, I was actually running wpa_supplicant again at the time of the crash. What I have now is very different behavior in two configurations on the same machine. The i386 kernel without SMP with most debug enabled and serial console. wpa_supplicant times out. If I restart is, the kernel oopses, every time in a different place. The x86_64 kernel with SMP and with very few debug options. wpa_supplicant connects. Killing and restarting wpa_supplicant doesn't cause any problems. In fact, wpa_supplicant reconnects quickly. I can even ping the station from the AP, but the packet loss is horrible. It appears that most loss is on the receiving side. I'll try to debug the problem when I have time. At least I'll try to find out if it's specific to the architecture or to another kernel option. Anyway, it's exciting that I could send first packets today! -- Regards, Pavel Roskin - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 9/9] net: vm deadlock avoidance core
On Wed, Jan 17, 2007 at 10:07:28AM +0100, Peter Zijlstra ([EMAIL PROTECTED]) wrote: > > You operate with 'current' in different contexts without any locks which > > looks racy and even is not allowed. What will be 'current' for > > netif_rx() case, which schedules softirq from hard irq context - > > ksoftirqd, why do you want to set its flags? > > I don't touch current in hardirq context, do I (if I did, that is indeed > a mistake)? > > In all other contexts, current is valid. Well, if you think that setting PF_MEMALLOC flag for keventd and ksoftirqd is valid, then probably yes... > > > > I meant that you can just mark process which created such socket as > > > > PF_MEMALLOC, and clone that flag on forks and other relatest calls > > > > without > > > > all that checks for 'current' in different places. > > > > > > Ah, thats the wrong level to think here, these processes never reach > > > user-space - nor should these sockets. > > > > You limit this just to send an ack? > > What about 'level-7' ack as you described in introduction? > > Take NFS, it does full data traffic in kernel. NFS case is exactly the situation, when you only need to generate an ACK. > > > Also, I only want the processing of the actual network packet to be able > > > to eat the reserves, not any other thing that might happen in that > > > context. > > > > > > And since network processing is mostly done in softirq context I must > > > mark these sections like I did. > > > > You artificially limit system to just add a reserve to generate one ack. > > For that purpose you do not need to have all those flags - just reseve > > some data in network core and use it when system is in OOM (or reclaim) > > for critical data pathes. > > How would that end up being different, I would have to replace all > allocations done in the full network processing path. > > This seems a much less invasive method, all the (allocation) code can > stay the way it is and use the normal allocation functions. Ack is only generated in one place in TCP. And acutally we are starting to talk about different approach - having separated allocator for network, which will be turned on on OOM (reclaim or at any other time). If you do not mind, I would likw to refresh a discussion about network tree allocator, which utilizes own pool of pages, performs self-defragmentation of the memeory, is very SMP friendly in that regard that it is per-cpu like slab and never free objects on different CPUs, so they always stay in the same cache. Among other goodies it allows to have full sending/receiving zero-copy. Here is a link: http://tservice.net.ru/~s0mbre/old/?section=projects&item=nta > > > > > > > + /* > > > > > > > +decrease window size.. > > > > > > > +tcp_enter_quickack_mode(sk); > > > > > > > + */ > > > > > > > > > > > > How does this decrease window size? > > > > > > Maybe ack scheduling would be better handled by > > > > > > inet_csk_schedule_ack() > > > > > > or just directly send an ack, which in turn requires allocation, > > > > > > which > > > > > > can be bound to this received frame processing... > > > > > > > > > > It doesn't, I thought that it might be a good idea doing that, but > > > > > never > > > > > got around to actually figuring out how to do it. > > > > > > > > tcp_send_ack()? > > > > > > > > > > does that shrink the window automagically? > > > > Yes, it updates window, but having ack generated in that place is > > actually very wrong. In that place system has not processed incoming > > packet yet, so it can not generate correct ACK for received frame at > > all. And it seems that the only purpose of the whole patchset is to > > generate that poor ack - reseve 2007 ack packets (MAX_TCP_HEADER) > > in system startup and reuse them when you are under memory pressure. > > Right, I suspected something like that; hence I wanted to just shrink > the window. Anyway, this is not a very important issue. tcp_enter_quickack_mode() does not update window, it allows to send ack immediately after packet has been processed, window can be changed in any way TCP state machine and congestion control want. -- Evgeniy Polyakov - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 9/9] net: vm deadlock avoidance core
On Thu, 2007-01-18 at 13:41 +0300, Evgeniy Polyakov wrote: > > > What about 'level-7' ack as you described in introduction? > > > > Take NFS, it does full data traffic in kernel. > > NFS case is exactly the situation, when you only need to generate an ACK. No it is not, it needs the full RPC response. > > > You artificially limit system to just add a reserve to generate one ack. > > > For that purpose you do not need to have all those flags - just reseve > > > some data in network core and use it when system is in OOM (or reclaim) > > > for critical data pathes. > > > > How would that end up being different, I would have to replace all > > allocations done in the full network processing path. > > > > This seems a much less invasive method, all the (allocation) code can > > stay the way it is and use the normal allocation functions. > And acutally we are starting to talk about different approach - having > separated allocator for network, which will be turned on on OOM (reclaim > or at any other time). I think we might be, I'm more talking about requirements on the allocator, while you seem to talk about implementations. Replacing the allocator, or splitting it in two based on a condition are all fine as long as they observe the requirements. The requirement I add is that there is a reserve nobody touches unless given express permission. You could implement this by modifying each reachable allocator call site and stick a branch in and use an alternate allocator when the normal route fails and we do have permission; much like: foo = kmalloc(size, gfp_mask); + if (!foo && special) +foo = my_alloc(size) And earlier versions of this work did something like that. But it litters the code quite badly and its quite easy to miss spots. There can be quite a few allocations in processing network data. Hence my work on integrating this into the regular memory allocators. FYI; 'special' evaluates to something like: !(gfp_mask & __GFP_NOMEMALLOC) && ((gfp_mask & __GFP_EMERGENCY) || (!in_irq() && (current->flags & PF_MEMALLOC))) > If you do not mind, I would likw to refresh a > discussion about network tree allocator, > which utilizes own pool of > pages, very high order pages, no? This means that you have to either allocate at boot time and cannot resize/add pools; which means you waste all that memory if the network load never comes near using the reserved amount. Or, you get into all the same trouble the hugepages folks are trying so very hard to solve. > performs self-defragmentation of the memeory, Does it move memory about? All it does is try to avoid fragmentation by policy - a problem impossible to solve in general; but can achieve good results in view of practical limitations on program behaviour. Does your policy work for the given workload? we'll see. Also, on what level, each level has both internal and external fragmentation. I can argue that having large immovable objects in memory adds to the fragmentation issues on the page-allocator level. > is very SMP > friendly in that regard that it is per-cpu like slab and never free > objects on different CPUs, so they always stay in the same cache. This makes it very hard to guarantee a reserve limit. (Not impossible, just more difficult) > Among other goodies it allows to have full sending/receiving zero-copy. That won't ever work unless you have page aligned objects, otherwise you cannot map them into user-space. Which seems to be at odds with your tight packing/reduce internal fragmentation goals. Zero-copy entails mapping the page the hardware writes the packet in into user-space, right? Since its impossible to predict to whoem the next packet is addressed the packets must be written (by hardware) to different pages. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH V2] bcm43xx: Fix failure to deliver PCI-E interrupts
One thing I forgot is that the LED will be on only if the switch is in the on position _AND_ the interface is UP. Yes I figured that out. (eventually) >>> dhcpcd eth1 <-low transmit power??? >> Most likely. How far are you from the AP? > > This is a timeout issue. I tried to apply the timeout patch which I > assume is for this and it didn't take. I have logs for both patch > applications below. To the distance issue my ap died so I am trying to > just connect to one next door. (with permission) The distance is about > 50ft. So this could be the main factor. This is why I said it might be > my fault. The thing is I can connect to it in windows most of the time > so I know it is in range. I will be going to school on Tuesday and > will report back then. They have a much better AP. With the current transmit levels of the 4311, 50 feet through at least one wall is probably too much. I have tried 3 more AP at much closer distances <9ft and <4ft and it still won't work. I ether get "Network Down" or a timeout. > Presario1600 bcm43xx # patch -p0 ... > Presario1600 bcm43xx # patch -p0 ... Those patches are already applied to your code. Thanks for the clarification, and assistance. -- http://www.coe.neu.edu/~efoss/ http://evanfoss.googlepages.com/ - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Please pull 'upstream' branch of wireless-2.6
On Tue, Jan 02, 2007 at 09:42:47PM -0500, John W. Linville wrote: > The following changes since commit fe5f8e2a1c5c040209c598a28e19c55f30e1040d: > Zhu Yi (1): > ipw2100: Fix dropping fragmented small packet problem > > are found in the git repository at: > > git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git > upstream > > Daniel Drake (5): > zd1211rw: Generic HMAC initialization > zd1211rw: 2 new ZD1211B device ID's > zd1211rw: Consistency for address space constants > zd1211rw: Remove addressing abstraction > zd1211rw: Add ID for Linksys WUSBF54G > > John W. Linville (1): > softmac: avoid assert in ieee80211softmac_wx_get_rate > > Kai Engert (1): > prism54: add ethtool -i interface > > Larry Finger (1): > bcm43xx: Interrogate hardware-enable switch and update LEDs > > Michael Buesch (1): > Update Prism54 MAINTAINERS entry Jeff, it looks like you didn't pull this one yet. I'm going to rebase my upstream branch on top of my additions to my upstream-fixes branch. So, just ignore this request and wait for the next round of pull requests. The next round is ready to go, except I can't login to master.kernel.org today. As soon as that is resolved, I'll push and send the pull requests. Thanks, John P.S. If you haven't pulled due to some objection, please send me a note to let me know why. -- John W. Linville [EMAIL PROTECTED] - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 1/12] L2 network namespace (v3): current network namespace operations
Dmitry Mishin <[EMAIL PROTECTED]> writes: > On Wednesday 17 January 2007 23:16, Eric W. Biederman wrote: >> Dmitry Mishin <[EMAIL PROTECTED]> writes: >> >> > Added functions and macros required to operate with network namespaces. >> > They are required in order to switch network namespace for incoming packets > and >> > to not extend current network interface by additional network namespace > argue. >> >> Is exec_net only used in interrupt context? > I tried to do so. > >> Or how do you ensure a sleeping function does not get called and the >> kernel process comes back on another cpu? > Seems that I forgot to remove it's usage at least in one place - in > clone_net_ns(). If you caught more, please, let me know. Sure. It was not clear from what I saw of the patch it was intended to be restricted to only interrupt context. So if figured I would ask if that was what you meant. Eric - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: TKIP encryption should allocate enough tailroom
On Wed, Jan 17, 2007 at 11:46:35AM -0500, Brandon Craig Rhodes wrote: > Having further reviewed my code, I have changed my mind; the > ieee80211_crypt_tkip routines are not designed to be responsible for > creating enough headroom and tailroom. The "hostap" driver should be > doing this. In fact, I now see that the "hostap" driver actually > attempts to create enough headroom and tailroom, but computes them > incorrectly. Even then, if ieee80211_tkip_encrypt() didn't produce debug output for the not enough space-case, should that be added to catch other potentially broken drivers? - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: TKIP encryption should allocate enough tailroom
Pekka Pietikainen <[EMAIL PROTECTED]> writes: > On Wed, Jan 17, 2007 at 11:46:35AM -0500, Brandon Craig Rhodes wrote: >> Having further reviewed my code, I have changed my mind; the >> ieee80211_crypt_tkip routines are not designed to be responsible for >> creating enough headroom and tailroom. > > Even then, if ieee80211_tkip_encrypt() didn't produce debug output > for the not enough space-case, should that be added to catch other > potentially broken drivers? I think your idea is an excellent one, and would have prevented my having to add a half-dozen printk()'s to the code myself to discover what was going on! I would be happy to submit such a patch myself, but am not sure what the local kernel conventions are regarding error messages - and the ieee80211_crypt_tkip.c functions seem wildly inconsistent with regard to debugging messages! In some circumstances, debug messages are always produced; in several others, net_ratelimit() is called to decided whether to print an error (but why in these cases and not others?); and in many cases, nothing is printed at all (is this because convention would dictate that the caller discover the error and print something out?). If I want to generate a patch that festoons the ieee80211 functions with informative error messages, what are the guidelines? -- Brandon Craig Rhodes [EMAIL PROTECTED] http://rhodesmill.org/brandon - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] Marvell Libertas 8388 802.11b/g USB driver (v2)
On Wed, 2007-01-17 at 23:09 +, Christoph Hellwig wrote: > On Wed, Jan 17, 2007 at 01:00:47PM -0500, Dan Williams wrote: > > Furthermore, I might add that the entire reason this part was chosen for > > OLPC was because it _is_ fullmac. > > > > To drive down power consumption, the OLPC puts the host CPU to sleep but > > keeps the 8388 powered on. That consumes around 400mW max. But it > > allows the 8388 to continue routing other laptops' packets over the mesh > > *while the host CPU is asleep*. > > We're not going to put a lot of junk into the kernel just because the OLPC > folks decide to do odd powermanagment schemes. Nor are we going to try to push highly OLPC-specific changes into this driver in a non-modular manner. This chip is also used in, for example, the X-Box 360 wireless dongle, and other non-OLPC people are working on adding support for the 8385 SDIO/CF variant. It would be arrogant of us to think that the driver would _just_ be useful for OLPC, and that's why nobody ever said that it was going to be an OLPC specific driver, not I, nor Marcelo. Give us some credit here before you start jumping all over something neither of us have said or implied. If we do make any OLPC specific changes (there are currently none [1]), they will be properly abstracted, conditionalized, and/or generalized. We will not be throwing random crap into this driver. Cheers, Dan [1] The code for the 802.11s mesh interface is in the driver, but that will/should automatically turn itself off if the firmware doesn't support that functionality. Furthermore, none of the mesh bits are OLPC-specific and may be used with any platform on which the driver runs. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] Marvell Libertas 8388 802.11b/g USB driver (v2)
On Wed, 2007-01-17 at 23:06 +, Christoph Hellwig wrote: > On Wed, Jan 17, 2007 at 04:42:04PM -0200, Marcelo Tosatti wrote: > > And using the Marvell provided firmware is a requirement for OLPC > > machines, where the CPU will be shut down but the chip+firmware will > > continue to forward packets in the mesh network (there are extreme power > > saving constraints on these machines). > > Well, than it probably doesn't go into mainline if you want to continue > doing this stupid layering violation. In the future we'll likely need some layering to support the 8385 SDIO/CF variant, but most likely not in way the USB support is currently excessively layered and abstracted. Dan - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Please pull 'upstream-fixes' branch of wireless-2.6
The following changes since commit a8b3485287731978899ced11f24628c927890e78: Linus Torvalds (1): Linux v2.6.20-rc5 are found in the git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git upstream-fixes Larry Finger (1): bcm43xx: Fix failure to deliver PCI-E interrupts drivers/net/wireless/bcm43xx/bcm43xx_main.c | 11 +++ 1 files changed, 7 insertions(+), 4 deletions(-) diff --git a/drivers/net/wireless/bcm43xx/bcm43xx_main.c b/drivers/net/wireless/bcm43xx/bcm43xx_main.c index 2ec2e5a..91b752e 100644 --- a/drivers/net/wireless/bcm43xx/bcm43xx_main.c +++ b/drivers/net/wireless/bcm43xx/bcm43xx_main.c @@ -2701,8 +2701,8 @@ static int bcm43xx_probe_cores(struct bcm43xx_private *bcm) sb_id_hi = bcm43xx_read32(bcm, BCM43xx_CIR_SB_ID_HI); /* extract core_id, core_rev, core_vendor */ - core_id = (sb_id_hi & 0xFFF0) >> 4; - core_rev = (sb_id_hi & 0xF); + core_id = (sb_id_hi & 0x8FF0) >> 4; + core_rev = ((sb_id_hi & 0xF) | ((sb_id_hi & 0x7000) >> 8)); core_vendor = (sb_id_hi & 0x) >> 16; dprintk(KERN_INFO PFX "Core %d: ID 0x%x, rev 0x%x, vendor 0x%x\n", @@ -2873,7 +2873,10 @@ static int bcm43xx_wireless_core_init(struct bcm43xx_private *bcm, sbimconfiglow = bcm43xx_read32(bcm, BCM43xx_CIR_SBIMCONFIGLOW); sbimconfiglow &= ~ BCM43xx_SBIMCONFIGLOW_REQUEST_TOUT_MASK; sbimconfiglow &= ~ BCM43xx_SBIMCONFIGLOW_SERVICE_TOUT_MASK; - sbimconfiglow |= 0x32; + if (bcm->bustype == BCM43xx_BUSTYPE_PCI) + sbimconfiglow |= 0x32; + else + sbimconfiglow |= 0x53; bcm43xx_write32(bcm, BCM43xx_CIR_SBIMCONFIGLOW, sbimconfiglow); } @@ -3077,7 +3080,7 @@ static int bcm43xx_setup_backplane_pci_connection(struct bcm43xx_private *bcm, if (err) goto out; - if (bcm->current_core->rev < 6 || + if (bcm->current_core->rev < 6 && bcm->current_core->id == BCM43xx_COREID_PCI) { value = bcm43xx_read32(bcm, BCM43xx_CIR_SBINTVEC); value |= (1 << backplane_flag_nr); -- John W. Linville [EMAIL PROTECTED] - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Please pull 'upstream' branch of wireless-2.6
The following changes since commit 10764889c6355cbb335cf0578ce12427475d1a65: Larry Finger (1): bcm43xx: Fix failure to deliver PCI-E interrupts are found in the git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git upstream Daniel Drake (6): zd1211rw: Generic HMAC initialization zd1211rw: 2 new ZD1211B device ID's zd1211rw: Consistency for address space constants zd1211rw: Remove addressing abstraction zd1211rw: Add ID for Linksys WUSBF54G zd1211rw: Add ID for ZyXEL ZyAIR G-220 v2 John W. Linville (1): softmac: avoid assert in ieee80211softmac_wx_get_rate Kai Engert (1): prism54: add ethtool -i interface Larry Finger (1): bcm43xx: Interrogate hardware-enable switch and update LEDs Michael Buesch (1): Update Prism54 MAINTAINERS entry Zhu Yi (1): ipw2200: add iwconfig rts/frag auto support MAINTAINERS |2 +- drivers/net/wireless/bcm43xx/bcm43xx.h|7 +- drivers/net/wireless/bcm43xx/bcm43xx_leds.c | 11 +- drivers/net/wireless/bcm43xx/bcm43xx_main.c | 36 -- drivers/net/wireless/bcm43xx/bcm43xx_radio.c |2 + drivers/net/wireless/bcm43xx/bcm43xx_radio.h | 16 +++ drivers/net/wireless/ipw2200.c|4 +- drivers/net/wireless/prism54/islpci_dev.c | 13 ++ drivers/net/wireless/prism54/islpci_dev.h |4 + drivers/net/wireless/prism54/islpci_hotplug.c |3 - drivers/net/wireless/zd1211rw/zd_chip.c | 126 ++-- drivers/net/wireless/zd1211rw/zd_chip.h | 158 ++--- drivers/net/wireless/zd1211rw/zd_def.h|2 + drivers/net/wireless/zd1211rw/zd_ieee80211.h |1 - drivers/net/wireless/zd1211rw/zd_rf.h |2 - drivers/net/wireless/zd1211rw/zd_types.h | 71 --- drivers/net/wireless/zd1211rw/zd_usb.c| 128 ++-- drivers/net/wireless/zd1211rw/zd_usb.h|6 +- net/ieee80211/softmac/ieee80211softmac_wx.c |6 + 19 files changed, 256 insertions(+), 342 deletions(-) delete mode 100644 drivers/net/wireless/zd1211rw/zd_types.h diff --git a/MAINTAINERS b/MAINTAINERS index 1b1491d..42b57cf 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -2654,7 +2654,7 @@ S:Supported PRISM54 WIRELESS DRIVER P: Prism54 Development Team -M: [EMAIL PROTECTED] +M: [EMAIL PROTECTED] L: netdev@vger.kernel.org W: http://prism54.org S: Maintained diff --git a/drivers/net/wireless/bcm43xx/bcm43xx.h b/drivers/net/wireless/bcm43xx/bcm43xx.h index 8286678..3a064de 100644 --- a/drivers/net/wireless/bcm43xx/bcm43xx.h +++ b/drivers/net/wireless/bcm43xx/bcm43xx.h @@ -352,6 +352,10 @@ #define BCM43xx_UCODEFLAG_UNKPACTRL0x0040 #define BCM43xx_UCODEFLAG_JAPAN0x0080 +/* Hardware Radio Enable masks */ +#define BCM43xx_MMIO_RADIO_HWENABLED_HI_MASK (1 << 16) +#define BCM43xx_MMIO_RADIO_HWENABLED_LO_MASK (1 << 4) + /* Generic-Interrupt reasons. */ #define BCM43xx_IRQ_READY (1 << 0) #define BCM43xx_IRQ_BEACON (1 << 1) @@ -758,7 +762,8 @@ struct bcm43xx_private { bad_frames_preempt:1, /* Use "Bad Frames Preemption" (default off) */ reg124_set_0x4:1, /* Some variable to keep track of IRQ stuff. */ short_preamble:1, /* TRUE, if short preamble is enabled. */ - firmware_norelease:1; /* Do not release the firmware. Used on suspend. */ + firmware_norelease:1, /* Do not release the firmware. Used on suspend. */ + radio_hw_enable:1; /* TRUE if radio is hardware enabled */ struct bcm43xx_stats stats; diff --git a/drivers/net/wireless/bcm43xx/bcm43xx_leds.c b/drivers/net/wireless/bcm43xx/bcm43xx_leds.c index 7d383a2..8f198be 100644 --- a/drivers/net/wireless/bcm43xx/bcm43xx_leds.c +++ b/drivers/net/wireless/bcm43xx/bcm43xx_leds.c @@ -26,6 +26,7 @@ */ #include "bcm43xx_leds.h" +#include "bcm43xx_radio.h" #include "bcm43xx.h" #include @@ -108,6 +109,7 @@ static void bcm43xx_led_init_hardcoded(struct bcm43xx_private *bcm, switch (led_index) { case 0: led->behaviour = BCM43xx_LED_ACTIVITY; + led->activelow = 1; if (bcm->board_vendor == PCI_VENDOR_ID_COMPAQ) led->behaviour = BCM43xx_LED_RADIO_ALL; break; @@ -199,20 +201,21 @@ void bcm43xx_leds_update(struct bcm43xx_private *bcm, int activity) turn_on = activity; break; case BCM43xx_LED_RADIO_ALL: - turn_on = radio->enabled; + turn_on = radio->enabled && bcm43xx_is_hw_radio_enabled(bcm); break; case BCM43xx_LED_RADIO_A: case BCM43xx_LED_BCM4303_2: - turn_on = (radio->enabled && ph
Re: [PATCH] [IPV6] fixed the size of the netlink message notified by inet6_rt_notify().
Hi, >> Wed, 17 Jan 2007 13:33:22 +0100 >> [Subject: Re: [PATCH] [IPV6] fixed the size of the netlink message notified >> by inet6_rt_notify().] >> Patrick McHardy <[EMAIL PROTECTED]> wrote... > Somewhat related: I have this patch for 2.6.21 to get rid of the > BUG_ON()s. I think the problem is the return value of inet6_rt_notify() is less than expected. -- Noriaki TAKAMIYA - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Please pull 'upstream-fixes' branch of wireless-2.6
John W. Linville wrote: The following changes since commit a8b3485287731978899ced11f24628c927890e78: Linus Torvalds (1): Linux v2.6.20-rc5 are found in the git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git upstream-fixes Larry Finger (1): bcm43xx: Fix failure to deliver PCI-E interrupts drivers/net/wireless/bcm43xx/bcm43xx_main.c | 11 +++ 1 files changed, 7 insertions(+), 4 deletions(-) pulled - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 0/8] partial resend: e1000 fixes and updates
Kok, Auke wrote: Hi, This patch series contains exclusively fixes for e1000. Some of these patches were already sent in december, but didn't make it into any usptream tree yet. Most importantly, it addresses two issues in the recently merged msi interrupt handler and dynamic itr code. A performance fix and some minor cleanups are also added. This brings the driver up to version 7.3.20-k2. The summary below lists all patches. Once that were previously acked are annotated with (*) These patches apply against netdev-2.6 #upstream-linus commit 77aab8bf22042d1658d4adbca8b71779e7f2d0ff. Please pull: git pull git://lost.foo-projects.org/~ahkok/git/netdev-2.6 upstream-linus Sigh. I /know/ I've told you this before, but let's review the branches again. NEVER EVER use branch 'upstream-linus'. That is for Linus only. There are three branches you should care about: master Vanilla upstream Linus tree, as of my last pull upstream-fixes netdev fixes being sent to Linus for -rc upstreamEverything else pending, but not yet merged The decision tree is as follows: * If there are no dependent e1000 changes, send patches against master (or torvalds/linux-2.6.git, if you prefer) * If there are dependent e1000 changes in #upstream-fixes, send patches against #upstream-fixes * If there are dependent e1000 changes in #upstream, send patches against #upstream In the current tree there are no e1000 changes, so please regenerate your tree against vanilla linux-2.6.git and resubmit. I ACK patches 1-8. Jeff - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 2.6.20 1/5] s2io: updates for s2io driver.
Ananda Raju wrote: Hello, List of changes in this patch: This patch adds two load parameters napi and ufo. Previously NAPI was compilation option with these changes wan enable disable NAPI using load parameter. Also we are introducing ufo load parameter to enable/disable ufo feature Signed-off-by: Sivakumar Subramani <[EMAIL PROTECTED]> OK, you're getting closer :) Problems that need correcting: 1) Your email subject line is a one-line summary of the patch. "s2io: updates for s2io driver" is useless, because it tells us nothing about the patch itself. When applied in a series, git log master..upstream-fixes | git shortlog will produce Ananda Raju (5): s2io: updates for s2io driver s2io: updates for s2io driver s2io: updates for s2io driver s2io: updates for s2io driver s2io: updates for s2io driver which clearly makes it impossible to distinguish between changesets. Please re-read Rule #1 of http://linux.yyz.us/patch-format.html Also, re-read Rule #2. Everything in your email body before the "---" terminator is copied DIRECTLY into the kernel changelog. As such, comments like "Hello," and "List of changes in this patch:" must be hand-edited out of your email, before applying the patch. Please fix these problems and resubmit. Jeff - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 1/3] myri10ge: make wc_fifo usage load-time tunable
Brice Goglin wrote: Under some circumstances, using WC without the WC fifo is faster. So we make it possible to tune wc_fifo with a module parameter. Signed-off-by: Brice Goglin <[EMAIL PROTECTED]> --- drivers/net/myri10ge/myri10ge.c |6 +- 1 file changed, 5 insertions(+), 1 deletion(-) applied 1-3 - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] Update ucc_geth.c for new workqueue structure
[EMAIL PROTECTED] wrote: From: Timur Tabi <[EMAIL PROTECTED]> The workqueue interface changed with David Howell's patch on 11/22/2006 (SHA 65f27f38446e1976cc98fd3004b110fedcddd189). Several drivers were updated with that patch to handle the new interface, but ucc_geth.c was not one of them. This patch updates ucc_geth.c to support the new model. A compiler warning in set_mac_addr() was also fixed. Signed-off-by: Timur Tabi <[EMAIL PROTECTED]> Signed-off-by: Kumar Gala <[EMAIL PROTECTED]> --- drivers/net/ucc_geth.c | 12 +++- 1 files changed, 7 insertions(+), 5 deletions(-) applied - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 8/12] net_device seq_file
Stephen Hemminger <[EMAIL PROTECTED]> writes: > On Wed, 17 Jan 2007 19:11:52 +0300 > Dmitry Mishin <[EMAIL PROTECTED]> wrote: > >> Library function to create a seq_file in proc filesystem, >> showing some information for each netdevice. >> This code is present in the kernel in about 10 instances, and >> all of them can be converted to using introduced library function. >> >> Signed-off-by: Andrey Savochkin <[EMAIL PROTECTED]> > > But the implementations can use static data rather than allocation, > also the long term desire is to get rid of supporting most of these proc > interfaces. While we do have the interfaces we need to support them, and not let them bit rot. That said so long as we have to individually register and unregister proc entries it is an active measure by the caller and they should be able to handle allocation and freeing of data structures as well. Eric - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 0/8] partial resend: e1000 fixes and updates
Jeff Garzik wrote: Kok, Auke wrote: Hi, This patch series contains exclusively fixes for e1000. Some of these patches were already sent in december, but didn't make it into any usptream tree yet. Most importantly, it addresses two issues in the recently merged msi interrupt handler and dynamic itr code. A performance fix and some minor cleanups are also added. This brings the driver up to version 7.3.20-k2. The summary below lists all patches. Once that were previously acked are annotated with (*) These patches apply against netdev-2.6 #upstream-linus commit 77aab8bf22042d1658d4adbca8b71779e7f2d0ff. Please pull: git pull git://lost.foo-projects.org/~ahkok/git/netdev-2.6 upstream-linus Sigh. I /know/ I've told you this before, but let's review the branches again. NEVER EVER use branch 'upstream-linus'. That is for Linus only. gah, sorry about that. I've moved it all over to #master-e1000 which applies against bf81b46482c0fa8ea638e409d39768ea92a6b0f0 ("Linux 2.6.20-rc4 --Linus Torvalds"). Please pull: git pull git://lost.foo-projects.org/~ahkok/git/netdev-2.6 master-e1000 Thanks, Auke * If there are no dependent e1000 changes, send patches against master (or torvalds/linux-2.6.git, if you prefer) * If there are dependent e1000 changes in #upstream-fixes, send patches against #upstream-fixes * If there are dependent e1000 changes in #upstream, send patches against #upstream *writes cheatsheet* - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH REPOST 1/2] NET: Accurate packet scheduling for ATM/ADSL (kernel)
On Thu, 2007-01-18 at 05:05 +0100, Patrick McHardy wrote: > > Yesterday I was chatting about this at LCA 2007, and > > it dawned on me that there is a problem with the dual > > RTAB/STAB approach. > > > > Currently the lookup in the kernel is > > time_to_transmit_a_packet = RTAB[packet_length_seen_by_kernel] > > > > As I understand it, you are proposing to change that to > > time_to_transmit_a_packet > > = RTAB[STAB[packet_length_seen_by_kernel]] > > = RTAB[packet_length_seen_on_the_wire] > > > > Given RTAB is the same in both cases the results of the > > calculation will be different (and ergo wrong in one case > > or the other). RTAB can't change and remain compatible > > with old kernels. Ergo this approach breaks backward > > compatibility. > > RTABs don't change, they continue to work as before. But when > an STAB is present the lookup is based on the STAB size mapping. > Neither one is wrong, RTABs calculate the transmission time based > only on the specified rate, RTABs + STABs calculate the > transmission time based on the rate, but include external overhead. No argument with "RTAB works as before". But aren't you proposing to feed it the accurate packet lengths calculated STAB? For example, if the VOIP IP datagram is 60 bytes, older kernels will index RTAB by (60 + ethernet_header_size), STAB kernels will index it by 159 bytes (3 ATM cells - say). If "tc" doesn't do a uname call or something, then it will send the same RTAB to both kernels. Obviously the results returned by the RTAB lookup will be different. If you aren't proposing to feed the RTAB lookup with the output of STAB, then I still don't understand why the current ATM patch isn't needed. Or are you proposing tc behave differently on different kernel versions. (I have no problem with that, but isn't it officially frowned upon?) - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
[PATCH] sis190: failure to set the MAC address from EEPROM
Fix from http://bugzilla.kernel.org/show_bug.cgi?id=7747 Signed-off-by: Andrew Morton <[EMAIL PROTECTED]> Cc: <[EMAIL PROTECTED]> Signed-off-by: Francois Romieu <[EMAIL PROTECTED]> --- drivers/net/sis190.c |2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/drivers/net/sis190.c b/drivers/net/sis190.c index b70ed79..45d91b1 100644 --- a/drivers/net/sis190.c +++ b/drivers/net/sis190.c @@ -1562,7 +1562,7 @@ static int __devinit sis190_get_mac_addr_from_eeprom(struct pci_dev *pdev, for (i = 0; i < MAC_ADDR_LEN / 2; i++) { __le16 w = sis190_read_eeprom(ioaddr, EEPROMMACAddr + i); - ((u16 *)dev->dev_addr)[0] = le16_to_cpu(w); + ((u16 *)dev->dev_addr)[i] = le16_to_cpu(w); } sis190_set_rgmii(tp, sis190_read_eeprom(ioaddr, EEPROMInfo)); -- 1.4.4.4 - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] Marvell Libertas 8388 802.11b/g USB driver (v2)
On Thu, Jan 18, 2007 at 10:41:45AM -0500, Dan Williams wrote: > In the future we'll likely need some layering to support the 8385 > SDIO/CF variant, but most likely not in way the USB support is currently > excessively layered and abstracted. Yeah. Let's summarize my unfortunately a bit too nasty comments and your more helpfull replies :-) This driver still needs a lot more work, both to cleanup a lot of crap and integreate it better with the wireless stack. And OLPC needs this is not going to be an excuse of it's own. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] Marvell Libertas 8388 802.11b/g USB driver (v2)
On 1/18/07, Christoph Hellwig <[EMAIL PROTECTED]> wrote: On Thu, Jan 18, 2007 at 10:41:45AM -0500, Dan Williams wrote: > In the future we'll likely need some layering to support the 8385 > SDIO/CF variant, but most likely not in way the USB support is currently > excessively layered and abstracted. Yeah. Let's summarize my unfortunately a bit too nasty comments and your more helpfull replies :-) This driver still needs a lot more work, both to cleanup a lot of crap and integrate it better with the wireless stack. And OLPC needs this is not going to be an excuse of it's own. The main feature of this chip is the on-chip support for 802.11s in firmware. What is the plan for integrating 802.11s into the existing wireless stacks? Does it make sense to do a softmac type 802.11s implementation first to figure out the right places to put the hooks for the 8388 hardware implementation? -- Jon Smirl [EMAIL PROTECTED] - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 0/12] L2 network namespace (v3)
In article <[EMAIL PROTECTED]> (at Wed, 17 Jan 2007 18:51:14 +0300), Dmitry Mishin <[EMAIL PROTECTED]> says: > === > L2 network namespaces > > The most straightforward concept of network virtualization is complete > separation of namespaces, covering device list, routing tables, netfilter > tables, socket hashes, and everything else. > > On input path, each packet is tagged with namespace right from the > place where it appears from a device, and is processed by each layer > in the context of this namespace. > Non-root namespaces communicate with the outside world in two ways: by > owning hardware devices, or receiving packets forwarded them by their parent > namespace via pass-through device. Can you handle multicast / broadcast and IPv6, which are very important? --yoshfuji - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 2/12] forcedeth: ring access
Ayaz Abdulla wrote: This patch modifys ring access by using pointers. This avoids computing the current index and avoids accessing the base address of the rings. Signed-Off-By: Ayaz Abdulla <[EMAIL PROTECTED]> ACK patches 1-9. Applied patches 1-4 successfully, patch #5 failed to apply. Please resend patches 5+ against netdev-2.6.git#upstream. Also, git-am (the git component used to apply yours and others' patches) complains that your patches have spaces prior to the tab indentations. Fixing that in patches 5+, or appending a final whitespace cleanup patch, would be appreciated. Jeff - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 10/12] forcedeth: tx max work
Ayaz Abdulla wrote: This patch adds a limit to how much tx work can be done in each iteration of tx processing. Signed-Off-By: Ayaz Abdulla <[EMAIL PROTECTED]> What about the "tail end" of the work, when the limit is reached? Remember that delaying the completion of TX's too long increases latency. It seems to me that this patch needs a timer or somesuch, to guarantee that TX completions are not delayed too long in the worst case. Jeff - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 11/12] forcedeth: statistics supported
Ayaz Abdulla wrote: This patch introduces hw statistics for older devices that supported it. It breaks up the counters supported into separate versions. Signed-Off-By: Ayaz Abdulla <[EMAIL PROTECTED]> ACK patches 11-12 - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 2/3] chelsio: NAPI speed improvement
Stephen Hemminger wrote: Speedup and cleanup the receive processing by eliminating the mmio read and a lock round trip. Signed-off-by: Stephen Hemminger <[EMAIL PROTECTED]> applied - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 3/3] chelsio: more rx speedup
Stephen Hemminger wrote: Cleanup receive processing some more: * do the reserve padding of skb during setup * don't pass constants to get_packet * do smart prefetch of skb * make copybreak a module parameter Signed-off-by: Stephen Hemminger <[EMAIL PROTECTED]> applied - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [2.6 patch] bonding.h: "extern inline" -> "static inline"
Adrian Bunk wrote: "extern inline" generates a warning with -Wmissing-prototypes and I'm currently working on getting the kernel cleaned up for adding this to the CFLAGS since it will help us to avoid a nasty class of runtime errors. If there are places that really need a forced inline, __always_inline would be the correct solution. Signed-off-by: Adrian Bunk <[EMAIL PROTECTED]> applied - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Please pull 'upstream' branch of wireless-2.6
John W. Linville wrote: The following changes since commit 10764889c6355cbb335cf0578ce12427475d1a65: Larry Finger (1): bcm43xx: Fix failure to deliver PCI-E interrupts are found in the git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6.git upstream ACK. Open question of parentage, though: I just rebased netdev-2.6.git#upstream. Is your wireless-2.6 affected by this rebase? If not, I will go ahead and pull. Jeff - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH REPOST 1/2] NET: Accurate packet scheduling for ATM/ADSL (kernel)
On Thu, 2007-01-18 at 12:37 +0100, Patrick McHardy wrote: > > Or are you proposing tc behave differently on different > > kernel versions. (I have no problem with that, but > > isn't it officially frowned upon?) > > Yes. There is no way you can make this work on old kernels, > nobody expects that. The important part is that everything > continues to work as before and that both old and new iproute > binaries work properly on both old and new kernels (new > iproute on old kernels without STABs obviously). I thought that some degree of compatibility was expected. At the very least the newest version of "tc" must work on _any_ kernel as least as well as the version it replaces did. I also though newer kernels should work older version of iproute2, albeit without the features added in the newer versions. Are you saying this is not so? - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] Marvell Libertas 8388 802.11b/g USB driver (v2)
On Thu, 2007-01-18 at 22:40 +, Christoph Hellwig wrote: > On Thu, Jan 18, 2007 at 10:41:45AM -0500, Dan Williams wrote: > > In the future we'll likely need some layering to support the 8385 > > SDIO/CF variant, but most likely not in way the USB support is currently > > excessively layered and abstracted. > > Yeah. Let's summarize my unfortunately a bit too nasty comments and > your more helpfull replies :-) > > This driver still needs a lot more work, both to cleanup a lot of crap > and integreate it better with the wireless stack. And OLPC needs this > is not going to be an excuse of it's own. Of course; the we-need-it-now-so-you-take-it-now attitude never gets vendors or anyone else anywhere, so there's no reason to expect it would work for OLPC either. OLPC does not need, nor does any of us who work on it ask for, special treatment. Stuff will go through the correct channels and will follow the normal kernel process. That's all we can expect, and that's all we ask. Cheers, Dan - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] Marvell Libertas 8388 802.11b/g USB driver (v2)
On Thu, 2007-01-18 at 17:54 -0500, Jon Smirl wrote: > On 1/18/07, Christoph Hellwig <[EMAIL PROTECTED]> wrote: > > On Thu, Jan 18, 2007 at 10:41:45AM -0500, Dan Williams wrote: > > > In the future we'll likely need some layering to support the 8385 > > > SDIO/CF variant, but most likely not in way the USB support is currently > > > excessively layered and abstracted. > > > > Yeah. Let's summarize my unfortunately a bit too nasty comments and > > your more helpfull replies :-) > > > > This driver still needs a lot more work, both to cleanup a lot of crap > > and integrate it better with the wireless stack. And OLPC needs this > > is not going to be an excuse of it's own. > > The main feature of this chip is the on-chip support for 802.11s in > firmware. What is the plan for integrating 802.11s into the existing > wireless stacks? Does it make sense to do a softmac type 802.11s > implementation first to figure out the right places to put the hooks > for the 8388 hardware implementation? I believe Javier Cordona (who will also be at the Linux Wireless Summit this weekend) is going to do a d80211-based implementation alongside the Libertas 8388 firmware and driver bits too. 802.11s networking in Linux is still quite immature, and we need to get people interested in a standard stack talking to each other. Dan - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] [IPV6] fixed the size of the netlink message notified by inet6_rt_notify().
In article <[EMAIL PROTECTED]> (at Fri, 19 Jan 2007 01:31:08 +0900 (JST)), Noriaki TAKAMIYA <[EMAIL PROTECTED]> says: > >> Wed, 17 Jan 2007 13:33:22 +0100 > >> [Subject: Re: [PATCH] [IPV6] fixed the size of the netlink message > >> notified by inet6_rt_notify().] > >> Patrick McHardy <[EMAIL PROTECTED]> wrote... > > > Somewhat related: I have this patch for 2.6.21 to get rid of the > > BUG_ON()s. > > I think the problem is the return value of inet6_rt_notify() is less > than expected. I guess Patrick meant that using "BUG_ON()" would be an overkill. --yoshfuji - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Possible ways of dealing with OOM conditions.
On Thu, Jan 18, 2007 at 01:18:44PM +0100, Peter Zijlstra ([EMAIL PROTECTED]) wrote: > > > How would that end up being different, I would have to replace all > > > allocations done in the full network processing path. > > > > > > This seems a much less invasive method, all the (allocation) code can > > > stay the way it is and use the normal allocation functions. > > > And acutally we are starting to talk about different approach - having > > separated allocator for network, which will be turned on on OOM (reclaim > > or at any other time). > > I think we might be, I'm more talking about requirements on the > allocator, while you seem to talk about implementations. > > Replacing the allocator, or splitting it in two based on a condition are > all fine as long as they observe the requirements. > > The requirement I add is that there is a reserve nobody touches unless > given express permission. > > You could implement this by modifying each reachable allocator call site > and stick a branch in and use an alternate allocator when the normal > route fails and we do have permission; much like: > >foo = kmalloc(size, gfp_mask); > + if (!foo && special) > +foo = my_alloc(size) Network is special in this regard, since it only has one allocation path (actually it has one cache for skb, and usual kmalloc, but they are called from only two functions). So it would become ptr = network_alloc(); and network_alloc() would be usual kmalloc or call for own allocator in case of deadlock. > And earlier versions of this work did something like that. But it > litters the code quite badly and its quite easy to miss spots. There can > be quite a few allocations in processing network data. > > Hence my work on integrating this into the regular memory allocators. > > FYI; 'special' evaluates to something like: > !(gfp_mask & __GFP_NOMEMALLOC) && > ((gfp_mask & __GFP_EMERGENCY) || >(!in_irq() && (current->flags & PF_MEMALLOC))) > > > > If you do not mind, I would likw to refresh a > > discussion about network tree allocator, > > > which utilizes own pool of > > pages, > > very high order pages, no? > > This means that you have to either allocate at boot time and cannot > resize/add pools; which means you waste all that memory if the network > load never comes near using the reserved amount. > > Or, you get into all the same trouble the hugepages folks are trying so > very hard to solve. It is configurable - by default it takes pool of 32k pages for allocations for jumbo-frames (e1000 requires such allocations for 9k frames unfortunately), without jumbo-frame support it works with pool of 0-order pages, which grows dynamically when needed. > > performs self-defragmentation of the memeory, > > Does it move memory about? It works in a page, not as pages - when neighbour regions are freed, they are combined into single one with bigger size - it would be extended to move pages around to combied them into bigger one though too, but network stack requires high-order allocations in extremely rare cases of broken design (Intel folks, sorry, but your hardware sucks in that regard - jumbo frame of 9k should not require 16k of mem plu network overhead). NTA also does not align buffers to the power of two - extremely significant win of that approach can be found on project's homepage with graps of failed allocations and state of the mem for different sizes of allocaions. Power-of-two overhead of SLAB is extremely high. > All it does is try to avoid fragmentation by policy - a problem > impossible to solve in general; but can achieve good results in view of > practical limitations on program behaviour. > > Does your policy work for the given workload? we'll see. > > Also, on what level, each level has both internal and external > fragmentation. I can argue that having large immovable objects in memory > adds to the fragmentation issues on the page-allocator level. NTA works with pages, not with contiguous memory, it reduces fragmentation inside pages, which can not be solved in SLAB, where objects from the same page can live in different caches and thus _never_ can be combined. Thus, the only soultuin for SLAB is copy, which is not a good one for big sizes and is just wrong for big pages. It is not about page moving and VM tricks, which are generally described as fragmentation avoidance technique, but about how fragmentation problem is solved in one page. > > is very SMP > > friendly in that regard that it is per-cpu like slab and never free > > objects on different CPUs, so they always stay in the same cache. > > This makes it very hard to guarantee a reserve limit. (Not impossible, > just more difficult) The whole pool of pages becomes reserve, since no one (and mainly VFS) can consume that reserve. > > Among other goodies it allows to have full sending/receiving zero-copy. > > That won't ever work unless you have page aligned objects, otherwise you > cannot map them into user-space. Whic
Re: Possible ways of dealing with OOM conditions.
On Thu, 2007-01-18 at 16:58 +0300, Evgeniy Polyakov wrote: > Network is special in this regard, since it only has one allocation path > (actually it has one cache for skb, and usual kmalloc, but they are > called from only two functions). > > So it would become > ptr = network_alloc(); > and network_alloc() would be usual kmalloc or call for own allocator in > case of deadlock. There is more to networking that skbs only, what about route cache, there is quite a lot of allocs in this fib_* stuff, IGMP etc... > > very high order pages, no? > > > > This means that you have to either allocate at boot time and cannot > > resize/add pools; which means you waste all that memory if the network > > load never comes near using the reserved amount. > > > > Or, you get into all the same trouble the hugepages folks are trying so > > very hard to solve. > > It is configurable - by default it takes pool of 32k pages for allocations for > jumbo-frames (e1000 requires such allocations for 9k frames > unfortunately), without jumbo-frame support it works with pool of 0-order > pages, which grows dynamically when needed. With 0-order pages, you can only fit 2 1500 byte packets in there, you could perhaps stick some small skb heads in there as well, but why bother, the waste isn't _that_ high. Esp if you would make a slab for 1500 mtu packets (5*1638 < 2*4096; and 1638 should be enough, right?) It would make sense to pack related objects into a page so you could free all together. > > > performs self-defragmentation of the memeory, > > > > Does it move memory about? > > It works in a page, not as pages - when neighbour regions are freed, > they are combined into single one with bigger size Yeah, that is not defragmentation, defragmentation is moving active regions about to create contiguous free space. What you do is free space coalescence. > but network stack requires high-order allocations in extremely rare > cases of broken design (Intel folks, sorry, but your hardware sucks in > that regard - jumbo frame of 9k should not require 16k of mem plu > network overhead). Well, if you have such hardware its not rare at all, But yeah that sucks. > NTA also does not align buffers to the power of two - extremely significant > win of that approach can be found on project's homepage with graps of > failed allocations and state of the mem for different sizes of > allocaions. Power-of-two overhead of SLAB is extremely high. Sure you can pack the page a little better(*), but I thought the main advantage was a speed increase. (*) memory is generally cheaper than engineering efforts, esp on this scale. The only advantage in the manual packing is that (with the fancy hardware stream engine mentioned below) you could ensure they are grouped together (then again, the hardware stream engine would, together with a SG-DMA engine, take care of that). > > > All it does is try to avoid fragmentation by policy - a problem > > impossible to solve in general; but can achieve good results in view of > > practical limitations on program behaviour. > > > > Does your policy work for the given workload? we'll see. > > > > Also, on what level, each level has both internal and external > > fragmentation. I can argue that having large immovable objects in memory > > adds to the fragmentation issues on the page-allocator level. > > NTA works with pages, not with contiguous memory, it reduces > fragmentation inside pages, which can not be solved in SLAB, where > objects from the same page can live in different caches and thus _never_ > can be combined. Thus, the only soultuin for SLAB is copy, which is not a > good one for big sizes and is just wrong for big pages. By allocating, and never returning the page to the page-allocator you've increased the fragmentation on the page-allocator level significantly. It will avoid a super page ever forming around that page. > It is not about page moving and VM tricks, which are generally described > as fragmentation avoidance technique, but about how fragmentation > problem is solved in one page. Short of defragmentation (move active regions about) fragmentation is an unsolved problem. For any heuristic there is a pattern that will defeat it. Luckily program allocation behaviour is usually very regular (or decomposable in well behaved groups). > > > is very SMP > > > friendly in that regard that it is per-cpu like slab and never free > > > objects on different CPUs, so they always stay in the same cache. > > > > This makes it very hard to guarantee a reserve limit. (Not impossible, > > just more difficult) > > The whole pool of pages becomes reserve, since no one (and mainly VFS) > can consume that reserve. Ah, but there you violate my requirement, any network allocation can claim the last bit of memory. The whole idea was that the reserve is explicitly managed. It not only needs protection from other users but also from itself. > > > Among other goodies it allows to have full sending/
Re: Possible ways of dealing with OOM conditions.
On Thu, Jan 18, 2007 at 04:10:52PM +0100, Peter Zijlstra ([EMAIL PROTECTED]) wrote: > On Thu, 2007-01-18 at 16:58 +0300, Evgeniy Polyakov wrote: > > > Network is special in this regard, since it only has one allocation path > > (actually it has one cache for skb, and usual kmalloc, but they are > > called from only two functions). > > > > So it would become > > ptr = network_alloc(); > > and network_alloc() would be usual kmalloc or call for own allocator in > > case of deadlock. > > There is more to networking that skbs only, what about route cache, > there is quite a lot of allocs in this fib_* stuff, IGMP etc... skbs are the most extensively used path. Actually the same is applied to route - dst_entries and rtable are allocated through own wrappers. > > > very high order pages, no? > > > > > > This means that you have to either allocate at boot time and cannot > > > resize/add pools; which means you waste all that memory if the network > > > load never comes near using the reserved amount. > > > > > > Or, you get into all the same trouble the hugepages folks are trying so > > > very hard to solve. > > > > It is configurable - by default it takes pool of 32k pages for allocations > > for > > jumbo-frames (e1000 requires such allocations for 9k frames > > unfortunately), without jumbo-frame support it works with pool of 0-order > > pages, which grows dynamically when needed. > > With 0-order pages, you can only fit 2 1500 byte packets in there, you > could perhaps stick some small skb heads in there as well, but why > bother, the waste isn't _that_ high. > > Esp if you would make a slab for 1500 mtu packets (5*1638 < 2*4096; and > 1638 should be enough, right?) > > It would make sense to pack related objects into a page so you could > free all together. With power-of-two allocation SLAB wastes 500 bytes for each 1500 MTU packet (roughly), it is actaly one ACK packet - and I hear it from person who develops a system, which is aimed to guarantee ACK allocation in OOM :) SLAB overhead is _very_ expensive for network - what if jumbo frame is used? It becomes incredible in that case, although modern NICs allows scatter-gather, which is aimed to fix the problem. Cache misses for small packet flow due to the fact, that the same data is allocated and freed and accessed on different CPUs will become an issue soon, not right now, since two-four core CPUs are not yet to be very popular and price for the cache miss is not _that_ high. > > > > performs self-defragmentation of the memeory, > > > > > > Does it move memory about? > > > > It works in a page, not as pages - when neighbour regions are freed, > > they are combined into single one with bigger size > > Yeah, that is not defragmentation, defragmentation is moving active > regions about to create contiguous free space. What you do is free space > coalescence. That is wrong definition just because no one developed different system. Defragmentation is a result of broken system. Existing design _does_not_ allow to have the situation when whole page belongs to the same cache after it was actively used, the same is applied to the situation when several pages, which create contiguous region, are used by different users, so people start develop VM tricks to move pages around so they would be placed near in address space. Do not fix the result, fix the reason. > > but network stack requires high-order allocations in extremely rare > > cases of broken design (Intel folks, sorry, but your hardware sucks in > > that regard - jumbo frame of 9k should not require 16k of mem plu > > network overhead). > > Well, if you have such hardware its not rare at all, But yeah that > sucks. They do a good jop developing different approaches to workaround that hardware 'feature', but this is still wrong situation. > > NTA also does not align buffers to the power of two - extremely significant > > win of that approach can be found on project's homepage with graps of > > failed allocations and state of the mem for different sizes of > > allocaions. Power-of-two overhead of SLAB is extremely high. > > Sure you can pack the page a little better(*), but I thought the main > advantage was a speed increase. > > (*) memory is generally cheaper than engineering efforts, esp on this > scale. The only advantage in the manual packing is that (with the fancy > hardware stream engine mentioned below) you could ensure they are > grouped together (then again, the hardware stream engine would, together > with a SG-DMA engine, take care of that). Extensoin way of doing things. That is wrong. > > > All it does is try to avoid fragmentation by policy - a problem > > > impossible to solve in general; but can achieve good results in view of > > > practical limitations on program behaviour. > > > > > > Does your policy work for the given workload? we'll see. > > > > > > Also, on what level, each level has both internal and external > > > fragmentation. I can argue that havi
Re: [PATCH 2.6.20-rc3]: 8139cp: Don't blindly enable interrupts
Chris Lalancette wrote: Francois Romieu wrote: Chris Lalancette <[EMAIL PROTECTED]> : [...] Thanks for the comments. While the patch you sent will help, there are still other places that will have problems. For example, in netpoll_send_skb, we call local_irq_save(flags), then call dev->hard_start_xmit(), and then call local_irq_restore(flags). This is a similar situation to what I described above; we will re-enable interrupts in cp_start_xmit(), when netpoll_send_skb doesn't expect that, and will probably run into issues. Is there a problem with changing cp_start_xmit to use the spin_lock_irqsave(), besides the extra instructions it needs? No. Given the history of locking in netpoll and the content of Documentation/networking/netdevices.txt, asking Herbert which rule(s) the code is supposed to follow seemed safer to me. You can forget my patch. Please resend your patch inlined to Jeff as described in http://linux.yyz.us/patch-format.html. Francois, Great. Resending mail, shortening subject to < 65 characters and inlining the patch. Thanks, Chris Lalancette Similar to this commit: http://kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=d15e9c4d9a75702b30e00cdf95c71c88e3f3f51e It's not safe in cp_start_xmit to blindly call spin_lock_irq and then spin_unlock_irq, since it may very well be the case that cp_start_xmit was called with interrupts already disabled (I came across this bug in the context of netdump in RedHat kernels, but the same issue holds, for example, in netconsole). Therefore, replace all instances of spin_lock_irq and spin_unlock_irq with spin_lock_irqsave and spin_unlock_irqrestore, respectively, in cp_start_xmit(). I tested this against a fully-virtualized Xen guest using netdump, which happens to use the 8139cp driver to talk to the emulated hardware. I don't have a real piece of 8139cp hardware to test on, so someone else will have to do that. Signed-off-by: Chris Lalancette <[EMAIL PROTECTED]> applied. In the future, please remove the quoted emails stuff, and anything else that does not belong in the kernel changelog. It must be hand-edited out, before using git-am to merge your patch into the kernel tree. Jeff - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Possible ways of dealing with OOM conditions.
On Thu, 2007-01-18 at 18:50 +0300, Evgeniy Polyakov wrote: > On Thu, Jan 18, 2007 at 04:10:52PM +0100, Peter Zijlstra ([EMAIL PROTECTED]) > wrote: > > On Thu, 2007-01-18 at 16:58 +0300, Evgeniy Polyakov wrote: > > > > > Network is special in this regard, since it only has one allocation path > > > (actually it has one cache for skb, and usual kmalloc, but they are > > > called from only two functions). > > > > > > So it would become > > > ptr = network_alloc(); > > > and network_alloc() would be usual kmalloc or call for own allocator in > > > case of deadlock. > > > > There is more to networking that skbs only, what about route cache, > > there is quite a lot of allocs in this fib_* stuff, IGMP etc... > > skbs are the most extensively used path. > Actually the same is applied to route - dst_entries and rtable are > allocated through own wrappers. Still, edit all places and perhaps forget one and make sure all new code doesn't forget about it, or pick a solution that covers everything. > With power-of-two allocation SLAB wastes 500 bytes for each 1500 MTU > packet (roughly), it is actaly one ACK packet - and I hear it from > person who develops a system, which is aimed to guarantee ACK > allocation in OOM :) I need full data traffic during OOM, not just a single ACK. > SLAB overhead is _very_ expensive for network - what if jumbo frame is > used? It becomes incredible in that case, although modern NICs allows > scatter-gather, which is aimed to fix the problem. Jumbo frames are fine if the hardware can do SG-DMA.. > Cache misses for small packet flow due to the fact, that the same data > is allocated and freed and accessed on different CPUs will become an > issue soon, not right now, since two-four core CPUs are not yet to be > very popular and price for the cache miss is not _that_ high. SGI does networking too, right? > > > > > performs self-defragmentation of the memeory, > > > > > > > > Does it move memory about? > > > > > > It works in a page, not as pages - when neighbour regions are freed, > > > they are combined into single one with bigger size > > > > Yeah, that is not defragmentation, defragmentation is moving active > > regions about to create contiguous free space. What you do is free space > > coalescence. > > That is wrong definition just because no one developed different system. > Defragmentation is a result of broken system. > > Existing design _does_not_ allow to have the situation when whole page > belongs to the same cache after it was actively used, the same is > applied to the situation when several pages, which create contiguous > region, are used by different users, so people start develop VM tricks > to move pages around so they would be placed near in address space. > > Do not fix the result, fix the reason. *plonk* 30+yrs of research ignored. > > > The whole pool of pages becomes reserve, since no one (and mainly VFS) > > > can consume that reserve. > > > > Ah, but there you violate my requirement, any network allocation can > > claim the last bit of memory. The whole idea was that the reserve is > > explicitly managed. > > > > It not only needs protection from other users but also from itself. > > Specifying some users as good and others as bad generally tends to very > bad behaviour. Your appwoach only covers some users, mine does not > differentiate between users, The kernel is special, right? It has priority over whatever user-land does. > but prevents system from such situation at all. I'm not seeing that, with your approach nobody stops the kernel from filling up the memory with user-space network traffic. swapping is not some random user process, its a fundamental kernel task, if this fails the machine is history. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Possible ways of dealing with OOM conditions.
On Thu, Jan 18, 2007 at 06:31:53PM +0100, Peter Zijlstra ([EMAIL PROTECTED]) wrote: > > skbs are the most extensively used path. > > Actually the same is applied to route - dst_entries and rtable are > > allocated through own wrappers. > > Still, edit all places and perhaps forget one and make sure all new code > doesn't forget about it, or pick a solution that covers everything. There is _one_ place for allocation of any kind of object. skb path has two places. > > With power-of-two allocation SLAB wastes 500 bytes for each 1500 MTU > > packet (roughly), it is actaly one ACK packet - and I hear it from > > person who develops a system, which is aimed to guarantee ACK > > allocation in OOM :) > > I need full data traffic during OOM, not just a single ACK. But your code exactly limit codepath to several allocaions, which must be ACK. You do not have enough reserve to support whole traffic. So the right solution, IMO, is to _prevent_ such situation, which means that allocation is not allowed to depend on external conditions like VFS. Actually my above sentences were about the case, when anly having different allocator, it is possible to dramatically change memory usage model, which supffers greatly from power-of-two allocations. OOM condition is one of the results which has big SLAB overhead among other roots. Actually all pathes which work with kmem_cache are safe against it, since kernel cache packs objects, but thos who uses raw kmalloc has problems. > > SLAB overhead is _very_ expensive for network - what if jumbo frame is > > used? It becomes incredible in that case, although modern NICs allows > > scatter-gather, which is aimed to fix the problem. > > Jumbo frames are fine if the hardware can do SG-DMA.. Notice word _IF_ in you sentence. e1000 for example can not (or it can, but driver is not developed for such scenario). > > Cache misses for small packet flow due to the fact, that the same data > > is allocated and freed and accessed on different CPUs will become an > > issue soon, not right now, since two-four core CPUs are not yet to be > > very popular and price for the cache miss is not _that_ high. > > SGI does networking too, right? Yep, Cristoph Lameter developed own allocator too. I agreee with you, that if that price is too high already, then it is a dditional sign to look into network tree allocator (yep, name is bad) again. > > That is wrong definition just because no one developed different system. > > Defragmentation is a result of broken system. > > > > Existing design _does_not_ allow to have the situation when whole page > > belongs to the same cache after it was actively used, the same is > > applied to the situation when several pages, which create contiguous > > region, are used by different users, so people start develop VM tricks > > to move pages around so they would be placed near in address space. > > > > Do not fix the result, fix the reason. > > *plonk* 30+yrs of research ignored. 30 years to develop SLAB allocator? In what universe that is all about? > > > > The whole pool of pages becomes reserve, since no one (and mainly VFS) > > > > can consume that reserve. > > > > > > Ah, but there you violate my requirement, any network allocation can > > > claim the last bit of memory. The whole idea was that the reserve is > > > explicitly managed. > > > > > > It not only needs protection from other users but also from itself. > > > > Specifying some users as good and others as bad generally tends to very > > bad behaviour. Your appwoach only covers some users, mine does not > > differentiate between users, > > The kernel is special, right? It has priority over whatever user-land > does. Kernel only does ACK generation and allocation for userspace. Kernel does not know that some of users are potentially good or bad, and if you will export this socket option to the userspace, everyone will think that his application is good enough to use reserve. So, for kernel-only side you just need to preallocate pool of packets and use them when system is in OOM (reclaim). For the long direction, new approach of memory allocaiton should be developed, and there are different works in that direction - NTA is one of them and not the only one, for the best resutlts it must be combined with vm-tricks defragmentation too. > > but prevents system from such situation at all. > > I'm not seeing that, with your approach nobody stops the kernel from > filling up the memory with user-space network traffic. > > swapping is not some random user process, its a fundamental kernel task, > if this fails the machine is history. You completely misses the point. The main goal is to 1. reduce fragmentation and/or enable self defragmentation (which is done in NTA), this also reduces memory usage. 2. perform correct recover steps in OOM - reduce memory usage, use different allocator and/or reserve (which is the case, where NTA can be used) 3. do not allow OOM condition - unfortunately it
Re: [PATCH 2.6.20-rc3]: 8139cp: Don't blindly enable interrupts
Jeff Garzik wrote: > Chris Lalancette wrote: > >> Francois Romieu wrote: >> >>> Chris Lalancette <[EMAIL PROTECTED]> : >>> [...] >>> >>> Thanks for the comments. While the patch you sent will help, there are still other places that will have problems. For example, in netpoll_send_skb, we call local_irq_save(flags), then call dev->hard_start_xmit(), and then call local_irq_restore(flags). This is a similar situation to what I described above; we will re-enable interrupts in cp_start_xmit(), when netpoll_send_skb doesn't expect that, and will probably run into issues. Is there a problem with changing cp_start_xmit to use the spin_lock_irqsave(), besides the extra instructions it needs? >>> >>> No. Given the history of locking in netpoll and the content of >>> Documentation/networking/netdevices.txt, asking Herbert which rule(s) >>> the code is supposed to follow seemed safer to me. >>> >>> You can forget my patch. >>> >>> Please resend your patch inlined to Jeff as described in >>> http://linux.yyz.us/patch-format.html. >>> >>> >>> >> Francois, >> Great. Resending mail, shortening subject to < 65 characters and >> inlining the patch. >> >> Thanks, >> Chris Lalancette >> >> Similar to this commit: >> >> http://kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=d15e9c4d9a75702b30e00cdf95c71c88e3f3f51e >> >> >> It's not safe in cp_start_xmit to blindly call spin_lock_irq and then >> spin_unlock_irq, since it may very well be the case that cp_start_xmit >> was called with interrupts already disabled (I came across this bug in >> the context of netdump in RedHat kernels, but the same issue holds, for >> example, in netconsole). Therefore, replace all instances of >> spin_lock_irq and spin_unlock_irq with spin_lock_irqsave and >> spin_unlock_irqrestore, respectively, in cp_start_xmit(). I tested this >> against a fully-virtualized Xen guest using netdump, which happens to >> use the 8139cp driver to talk to the emulated hardware. I don't have a >> real piece of 8139cp hardware to test on, so someone else will have to >> do that. >> >> Signed-off-by: Chris Lalancette <[EMAIL PROTECTED]> > > > applied. > > In the future, please remove the quoted emails stuff, and anything > else that does not belong in the kernel changelog. It must be > hand-edited out, before using git-am to merge your patch into the > kernel tree. > > Jeff > > > Jeff, Ah, I see. Noted. Thanks. Chris Lalancette - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
[2.6 patch] drivers/net/irda/vlsi_ir.{h,c}: remove kernel 2.4 code
This patch removes kernel 2.4 compatibility code. Signed-off-by: Adrian Bunk <[EMAIL PROTECTED]> --- drivers/net/irda/vlsi_ir.c | 16 drivers/net/irda/vlsi_ir.h | 33 - 2 files changed, 8 insertions(+), 41 deletions(-) --- linux-2.6.20-rc4-mm1/drivers/net/irda/vlsi_ir.h.old 2007-01-18 21:50:43.0 +0100 +++ linux-2.6.20-rc4-mm1/drivers/net/irda/vlsi_ir.h 2007-01-18 21:53:54.0 +0100 @@ -41,39 +41,6 @@ #define PCI_CLASS_SUBCLASS_MASK0x #endif -/* in recent 2.5 interrupt handlers have non-void return value */ -#ifndef IRQ_RETVAL -typedef void irqreturn_t; -#define IRQ_NONE -#define IRQ_HANDLED -#define IRQ_RETVAL(x) -#endif - -/* some stuff need to check kernelversion. Not all 2.5 stuff was present - * in early 2.5.x - the test is merely to separate 2.4 from 2.5 - */ -#include - -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0) - -/* PDE() introduced in 2.5.4 */ -#ifdef CONFIG_PROC_FS -#define PDE(inode) ((inode)->i_private) -#endif - -/* irda crc16 calculation exported in 2.5.42 */ -#define irda_calc_crc16(fcs,buf,len) (GOOD_FCS) - -/* we use this for unified pci device name access */ -#define PCIDEV_NAME(pdev) ((pdev)->name) - -#else /* 2.5 or later */ - -/* whatever we get from the associated struct device - bus:slot:dev.fn id */ -#define PCIDEV_NAME(pdev) (pci_name(pdev)) - -#endif - /* */ /* non-standard PCI registers */ --- linux-2.6.20-rc4-mm1/drivers/net/irda/vlsi_ir.c.old 2007-01-18 21:53:58.0 +0100 +++ linux-2.6.20-rc4-mm1/drivers/net/irda/vlsi_ir.c 2007-01-18 21:54:56.0 +0100 @@ -166,7 +166,7 @@ unsigned i; seq_printf(seq, "\n%s (vid/did: %04x/%04x)\n", - PCIDEV_NAME(pdev), (int)pdev->vendor, (int)pdev->device); + pci_name(pdev), (int)pdev->vendor, (int)pdev->device); seq_printf(seq, "pci-power-state: %u\n", (unsigned) pdev->current_state); seq_printf(seq, "resources: irq=%u / io=0x%04x / dma_mask=0x%016Lx\n", pdev->irq, (unsigned)pci_resource_start(pdev, 0), (unsigned long long)pdev->dma_mask); @@ -1401,7 +1401,7 @@ if (vlsi_start_hw(idev)) IRDA_ERROR("%s: failed to restart hw - %s(%s) unusable!\n", - __FUNCTION__, PCIDEV_NAME(idev->pdev), ndev->name); + __FUNCTION__, pci_name(idev->pdev), ndev->name); else netif_start_queue(ndev); } @@ -1643,7 +1643,7 @@ pdev->current_state = 0; /* hw must be running now */ IRDA_MESSAGE("%s: IrDA PCI controller %s detected\n", -drivername, PCIDEV_NAME(pdev)); +drivername, pci_name(pdev)); if ( !pci_resource_start(pdev,0) || !(pci_resource_flags(pdev,0) & IORESOURCE_IO) ) { @@ -1728,7 +1728,7 @@ pci_set_drvdata(pdev, NULL); - IRDA_MESSAGE("%s: %s removed\n", drivername, PCIDEV_NAME(pdev)); + IRDA_MESSAGE("%s: %s removed\n", drivername, pci_name(pdev)); } #ifdef CONFIG_PM @@ -1748,7 +1748,7 @@ if (!ndev) { IRDA_ERROR("%s - %s: no netdevice \n", - __FUNCTION__, PCIDEV_NAME(pdev)); + __FUNCTION__, pci_name(pdev)); return 0; } idev = ndev->priv; @@ -1759,7 +1759,7 @@ pdev->current_state = state.event; } else - IRDA_ERROR("%s - %s: invalid suspend request %u -> %u\n", __FUNCTION__, PCIDEV_NAME(pdev), pdev->current_state, state.event); + IRDA_ERROR("%s - %s: invalid suspend request %u -> %u\n", __FUNCTION__, pci_name(pdev), pdev->current_state, state.event); up(&idev->sem); return 0; } @@ -1787,7 +1787,7 @@ if (!ndev) { IRDA_ERROR("%s - %s: no netdevice \n", - __FUNCTION__, PCIDEV_NAME(pdev)); + __FUNCTION__, pci_name(pdev)); return 0; } idev = ndev->priv; @@ -1795,7 +1795,7 @@ if (pdev->current_state == 0) { up(&idev->sem); IRDA_WARNING("%s - %s: already resumed\n", -__FUNCTION__, PCIDEV_NAME(pdev)); +__FUNCTION__, pci_name(pdev)); return 0; } - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
[BUG] 2.6.20-rc4-mm1: Panic in e1000_write_vfta_82543()
I get following panic on 2.6.20-rc4-mm1 on a 2-cpu AMD Opteron system. Same basic config file seems to work with 2.6.20-rc2-mm1 on this same system. Have not tried -rc3-mm1 yet. Attached are config file and "lspci -vv" output. Let me know if you need more info. Suka --- [ 168.925840] Freeing unused kernel memory: 320k freed * INIT: version 2.86 booting * Starting basic networking... [ ok ] * Starting kernel event manager... [ ok ] * Loading hardware drivers... [ ok ] * Starting PCMCIA services... * PCMCIA not present * Loading manual drivers... [ ok ] [ 171.575122] Unable to handle kernel paging request at c20100ec55fc RIP: [ 171.584632] [] e1000_write_vfta_82543+0x58/0xd0 [ 171.614833] PGD 114e067 PUD 0 [ 171.633943] Oops: [1] PREEMPT SMP [ 171.654614] last sysfs file: /devices/pci:00/:00:06.0/:01:06.0/host0/target0:0:0/0:0:0:0/rev [ 171.698158] CPU 1 [ 171.698161] Modules linked in: [ 171.698164] Pid: 2173, comm: ifconfig Not tainted 2.6.20-rc4-mm1 #1 * Checking root[ 171.698166] RIP: 0010:[] [] e1000_write_vfta_82543+0x58/0xd0 file system...[ 171.698171] RSP: 0018:81007dfc5cf8 EFLAGS: 00010286 [ 171.698174] RAX: c20100ec55fc RBX: 81003e4c8948 RCX: /dev/shm/root: clean, 662056/2443200 files, 3121107/4883752 bloc[ 171.698176] RDX: RSI: RDI: 81003e4c8948 ks [ ok ] [ 171.698178] RBP: R08: 0010 R09: [ 171.698181] R10: 0002 R11: 804a9800 R12: c2ec [ 171.698183] R13: fffc R14: R15: 81003e4c8000 * Setting up LVM Volume Groups...[ 171.698186] FS: 2ab17ce3f6d0() GS:81007f0b0bc0() knlGS: [ 171.698188] CS: 0010 DS: ES: CR0: 8005003b [ 171.698191] CR2: c20100ec55fc CR3: 7d828000 CR4: 06e0 [ 171.698194] Process ifconfig (pid: 2173, threadinfo 81007dfc4000, task 81007dd957e0) [ 171.698196] Stack: 8000 81003e4c8680 c2ec 81003e4c8000 [ 171.698202] 7fff2dfba960 804c103d 81003e4c8680 [ 171.698205] 81003e4c8680 804c3195 [ 171.698208] Call Trace: [ 171.698216] [] e1000_vlan_rx_register+0x1dd/0x210 [ 171.698219] [] e1000_up+0x35/0x4b0 [ 171.698222] [] e1000_open+0x74/0x100 [ 171.698227] [] dev_open+0x3e/0xa0 [ 171.698230] [] dev_change_flags+0x6f/0x160 [ 171.698234] [] devinet_ioctl+0x2d4/0x6e0 [ 171.698238] [] __up_read+0x21/0xb0 [ 171.698243] [] sock_ioctl+0x1fc/0x230 [ 171.698247] [] do_ioctl+0x2f/0xa0 [ 171.698249] [] vfs_ioctl+0x2bb/0x2f0 [ 171.698252] [] sys_ioctl+0x49/0x80 [ 171.698256] [] error_exit+0x0/0x84 [ 171.698259] [] system_call+0x7e/0x83 [ 171.698261] [ 171.698262] [ 171.698262] Code: 44 8b 20 e8 30 7e 00 00 83 bb 94 01 00 00 03 75 3c 83 e5 01 [ 171.698268] RIP [] e1000_write_vfta_82543+0x58/0xd0 [ 171.698273] RSP [ 171.698274] CR2: c20100ec55fc [ 171.698276] <6>EXT3 FS on sda1, internal journal # # Automatically generated make config: don't edit # Linux kernel version: 2.6.20-rc4-mm1 # Thu Jan 18 15:00:14 2007 # CONFIG_X86_64=y CONFIG_64BIT=y CONFIG_X86=y CONFIG_GENERIC_TIME=y CONFIG_GENERIC_TIME_VSYSCALL=y CONFIG_ZONE_DMA32=y CONFIG_LOCKDEP_SUPPORT=y CONFIG_STACKTRACE_SUPPORT=y CONFIG_SEMAPHORE_SLEEPERS=y CONFIG_MMU=y CONFIG_ZONE_DMA=y CONFIG_RWSEM_GENERIC_SPINLOCK=y CONFIG_GENERIC_HWEIGHT=y CONFIG_GENERIC_CALIBRATE_DELAY=y CONFIG_X86_CMPXCHG=y CONFIG_EARLY_PRINTK=y CONFIG_GENERIC_ISA_DMA=y CONFIG_GENERIC_IOMAP=y CONFIG_ARCH_MAY_HAVE_PC_FDC=y CONFIG_ARCH_POPULATES_NODE_MAP=y CONFIG_DMI=y CONFIG_AUDIT_ARCH=y CONFIG_GENERIC_BUG=y # CONFIG_ARCH_HAS_ILOG2_U32 is not set # CONFIG_ARCH_HAS_ILOG2_U64 is not set CONFIG_DEFCONFIG_LIST="/lib/modules/$UNAME_RELEASE/.config" # # Code maturity level options # CONFIG_EXPERIMENTAL=y CONFIG_LOCK_KERNEL=y CONFIG_INIT_ENV_ARG_LIMIT=32 # # General setup # CONFIG_LOCALVERSION="" # CONFIG_LOCALVERSION_AUTO is not set CONFIG_SWAP=y # CONFIG_SWAP_PREFETCH is not set CONFIG_SYSVIPC=y # CONFIG_IPC_NS is not set CONFIG_POSIX_MQUEUE=y CONFIG_BSD_PROCESS_ACCT=y CONFIG_BSD_PROCESS_ACCT_V3=y # CONFIG_TASKSTATS is not set # CONFIG_UTS_NS is not set CONFIG_USER_NS=y CONFIG_AUDIT=y # CONFIG_AUDITSYSCALL is not set CONFIG_IKCONFIG=y CONFIG_IKCONFIG_PROC=y CONFIG_CPUSETS=y CONFIG_SYSFS_DEPRECATED=y # CONFIG_RELAY is not set CONFIG_INITRAMFS_SOURCE="" # CONFIG_CC_OPTIMIZE_FOR_SIZE is not set CONFIG_SYSCTL=y # CONFIG_EMBEDDED is not set CONFIG_UID16=y CONFIG_SYSCTL_SYSCALL=y CONFIG_KALLSYMS=y # CONFIG_KALLSYMS_ALL is not set # CONFIG_KALLSYMS_EXTRA_PASS is not set CONFIG_HOTPLUG=y
Re: [BUG] 2.6.20-rc4-mm1: Panic in e1000_write_vfta_82543()
Auke Kok [EMAIL PROTECTED] wrote: | Sukadev Bhattiprolu wrote: | >I get following panic on 2.6.20-rc4-mm1 on a 2-cpu AMD Opteron system. | > | >Same basic config file seems to work with 2.6.20-rc2-mm1 on this same | >system. Have not tried -rc3-mm1 yet. | > | >Attached are config file and "lspci -vv" output. Let me know if you need | >more info. | > | >Suka | > | >--- | > | >[ 168.925840] Freeing unused kernel memory: 320k freed | > * INIT: version 2.86 booting | > * Starting basic networking... [ | > ok ] | > * Starting kernel event manager... [ | > ok ] | > * Loading hardware drivers... [ | > ok ] | > * Starting PCMCIA services... * PCMCIA not present | > * Loading manual drivers... [ | > ok ] | >[ 171.575122] Unable to handle kernel paging request at c20100ec55fc | >RIP: | >[ 171.584632] [] e1000_write_vfta_82543+0x58/0xd0 ... | >[ 171.698208] Call Trace: | >[ 171.698216] [] e1000_vlan_rx_register+0x1dd/0x210 | >[ 171.698219] [] e1000_up+0x35/0x4b0 | >[ 171.698222] [] e1000_open+0x74/0x100 | >[ 171.698227] [] dev_open+0x3e/0xa0 | >[ 171.698230] [] dev_change_flags+0x6f/0x160 | >[ 171.698234] [] devinet_ioctl+0x2d4/0x6e0 | >[ 171.698238] [] __up_read+0x21/0xb0 | >[ 171.698243] [] sock_ioctl+0x1fc/0x230 | >[ 171.698247] [] do_ioctl+0x2f/0xa0 | >[ 171.698249] [] vfs_ioctl+0x2bb/0x2f0 | >[ 171.698252] [] sys_ioctl+0x49/0x80 | >[ 171.698256] [] error_exit+0x0/0x84 | >[ 171.698259] [] system_call+0x7e/0x83 | >[ 171.698261] | >[ 171.698262] | >[ 171.698262] Code: 44 8b 20 e8 30 7e 00 00 83 bb 94 01 00 00 03 75 3c 83 | >e5 01 | >[ 171.698268] RIP [] e1000_write_vfta_82543+0x58/0xd0 | >[ 171.698273] RSP | >[ 171.698274] CR2: c20100ec55fc | >[ 171.698276] <6>EXT3 FS on sda1, internal journal | | Hi, | | I believe this is one of the bugs that is fixed in the patch that I sent | monday. Please | try again with the patch applied to your tree and re-test. Thanks. I didn't | see Andrew | merge the patch yet. | | see: http://lkml.org/lkml/2007/1/16/226 | | I can mail the patch if you can't find it. Just ping me privately. Yep. Seems to fix the crash. Thanks ! Suka | | Cheers, | | Auke - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [BUG] 2.6.20-rc4-mm1: Panic in e1000_write_vfta_82543()
Sukadev Bhattiprolu wrote: I get following panic on 2.6.20-rc4-mm1 on a 2-cpu AMD Opteron system. Same basic config file seems to work with 2.6.20-rc2-mm1 on this same system. Have not tried -rc3-mm1 yet. Attached are config file and "lspci -vv" output. Let me know if you need more info. Suka --- [ 168.925840] Freeing unused kernel memory: 320k freed * INIT: version 2.86 booting * Starting basic networking... [ ok ] * Starting kernel event manager... [ ok ] * Loading hardware drivers... [ ok ] * Starting PCMCIA services... * PCMCIA not present * Loading manual drivers... [ ok ] [ 171.575122] Unable to handle kernel paging request at c20100ec55fc RIP: [ 171.584632] [] e1000_write_vfta_82543+0x58/0xd0 [ 171.614833] PGD 114e067 PUD 0 [ 171.633943] Oops: [1] PREEMPT SMP [ 171.654614] last sysfs file: /devices/pci:00/:00:06.0/:01:06.0/host0/target0:0:0/0:0:0:0/rev [ 171.698158] CPU 1 [ 171.698161] Modules linked in: [ 171.698164] Pid: 2173, comm: ifconfig Not tainted 2.6.20-rc4-mm1 #1 * Checking root[ 171.698166] RIP: 0010:[] [] e1000_write_vfta_82543+0x58/0xd0 file system...[ 171.698171] RSP: 0018:81007dfc5cf8 EFLAGS: 00010286 [ 171.698174] RAX: c20100ec55fc RBX: 81003e4c8948 RCX: /dev/shm/root: clean, 662056/2443200 files, 3121107/4883752 bloc[ 171.698176] RDX: RSI: RDI: 81003e4c8948 ks [ ok ] [ 171.698178] RBP: R08: 0010 R09: [ 171.698181] R10: 0002 R11: 804a9800 R12: c2ec [ 171.698183] R13: fffc R14: R15: 81003e4c8000 * Setting up LVM Volume Groups...[ 171.698186] FS: 2ab17ce3f6d0() GS:81007f0b0bc0() knlGS: [ 171.698188] CS: 0010 DS: ES: CR0: 8005003b [ 171.698191] CR2: c20100ec55fc CR3: 7d828000 CR4: 06e0 [ 171.698194] Process ifconfig (pid: 2173, threadinfo 81007dfc4000, task 81007dd957e0) [ 171.698196] Stack: 8000 81003e4c8680 c2ec 81003e4c8000 [ 171.698202] 7fff2dfba960 804c103d 81003e4c8680 [ 171.698205] 81003e4c8680 804c3195 [ 171.698208] Call Trace: [ 171.698216] [] e1000_vlan_rx_register+0x1dd/0x210 [ 171.698219] [] e1000_up+0x35/0x4b0 [ 171.698222] [] e1000_open+0x74/0x100 [ 171.698227] [] dev_open+0x3e/0xa0 [ 171.698230] [] dev_change_flags+0x6f/0x160 [ 171.698234] [] devinet_ioctl+0x2d4/0x6e0 [ 171.698238] [] __up_read+0x21/0xb0 [ 171.698243] [] sock_ioctl+0x1fc/0x230 [ 171.698247] [] do_ioctl+0x2f/0xa0 [ 171.698249] [] vfs_ioctl+0x2bb/0x2f0 [ 171.698252] [] sys_ioctl+0x49/0x80 [ 171.698256] [] error_exit+0x0/0x84 [ 171.698259] [] system_call+0x7e/0x83 [ 171.698261] [ 171.698262] [ 171.698262] Code: 44 8b 20 e8 30 7e 00 00 83 bb 94 01 00 00 03 75 3c 83 e5 01 [ 171.698268] RIP [] e1000_write_vfta_82543+0x58/0xd0 [ 171.698273] RSP [ 171.698274] CR2: c20100ec55fc [ 171.698276] <6>EXT3 FS on sda1, internal journal Hi, I believe this is one of the bugs that is fixed in the patch that I sent monday. Please try again with the patch applied to your tree and re-test. Thanks. I didn't see Andrew merge the patch yet. see: http://lkml.org/lkml/2007/1/16/226 I can mail the patch if you can't find it. Just ping me privately. Cheers, Auke - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 4/4] atl1: Ancillary C files for Attansic L1 driver
Jay Cliburn wrote: +static u32 atl1_get_tx_csum(struct net_device *netdev) +{ + return (netdev->features & NETIF_F_HW_CSUM) != 0; +} + +static int atl1_set_tx_csum(struct net_device *netdev, u32 data) +{ + if (data) + netdev->features |= NETIF_F_HW_CSUM; + else + netdev->features &= ~NETIF_F_HW_CSUM; + + return 0; +} + +static int atl1_set_tso(struct net_device *netdev, u32 data) +{ + if (data) + netdev->features |= NETIF_F_TSO; + else + netdev->features &= ~NETIF_F_TSO; + return 0; +} There should be generic functions covering this. +static u32 ether_crc_le(int length, unsigned char *data) +{ + u32 crc = ~0; /* Initial value. */ + while (--length >= 0) { + unsigned char current_octet = *data++; + int bit; + for (bit = 8; --bit >= 0; current_octet >>= 1) { + if ((crc ^ current_octet) & 1) { + crc >>= 1; + crc ^= 0xedb88320; + } else + crc >>= 1; + } + } + return ~crc; +} this duplicates a library function +/** + * Reset the transmit and receive units; mask and clear all interrupts. + * hw - Struct containing variables accessed by shared code + * return : ATL1_SUCCESS or idle status (if error) + **/ +s32 atl1_reset_hw(struct atl1_hw * hw) +{ + u32 icr; + u16 pci_cfg_cmd_word; + int i; + + /* Workaround for PCI problem when BIOS sets MMRBC incorrectly. */ + atl1_read_pci_cfg(hw, PCI_REG_COMMAND, &pci_cfg_cmd_word); + if ((pci_cfg_cmd_word & +(CMD_IO_SPACE | CMD_MEMORY_SPACE | CMD_BUS_MASTER)) + != (CMD_IO_SPACE | CMD_MEMORY_SPACE | CMD_BUS_MASTER)) { + pci_cfg_cmd_word |= + (CMD_IO_SPACE | CMD_MEMORY_SPACE | CMD_BUS_MASTER); + atl1_write_pci_cfg(hw, PCI_REG_COMMAND, &pci_cfg_cmd_word); + } This duplicates bits set by pci_enable_device() and pci_set_master() + /** + * Clear Interrupt mask to stop board from generating + * interrupts & Clear any pending interrupt events + **/ + /** +* atl1_write32(hw, REG_IMR, 0); +* atl1_write32(hw, REG_ISR, 0x); +**/ + + /** +* Issue Soft Reset to the MAC. This will reset the chip's +* transmit, receive, DMA. It will not effect +* the current PCI configuration. The global reset bit is self- +* clearing, and should clear within a microsecond. +**/ + /*atl1_write32(hw, REG_MASTER_CTRL, MASTER_CTRL_SOFT_RST);*/ + atl1_write32(hw, REG_MASTER_CTRL, MASTER_CTRL_SOFT_RST); + wmb(); PCI posting (need to read a register before delaying to guarantee flush; probably makes wmb superfluous) + atl1_write16(hw, REG_GPHY_ENABLE, 1); + + msec_delay(1); /* delay about 1ms */ ditto + /* Wait at least 10ms for All module to be Idle */ + for (i = 0; i < 10; i++) { + icr = atl1_read32(hw, REG_IDLE_STATUS); + if (!icr) + break; + msec_delay(1); /* delay 1 ms */ + cpu_relax();/* FIXME: is this still the right way to do this? */ + } + + if (icr) + return icr; + + return ATL1_SUCCESS; +} + +static inline bool atl1_eth_address_valid(u8 * p_addr) +{ + /* Invalid PermanentAddress ? */ + if (((p_addr[0] == 0) && +(p_addr[1] == 0) && +(p_addr[2] == 0) && +(p_addr[3] == 0) && (p_addr[4] == 0) && (p_addr[5] == 0) + ) || (p_addr[0] & 1)) + /* Multicast address or Broadcast Address */ + return false; + + return true; look at is_valid_ether_addr() lib function + if (atl1_get_permanent_address(hw)) { + hw->perm_mac_addr[0] = 0x00; + hw->perm_mac_addr[1] = 0x13; + hw->perm_mac_addr[2] = 0x74; + hw->perm_mac_addr[3] = 0x00; + hw->perm_mac_addr[4] = 0x5c; + hw->perm_mac_addr[5] = 0x38; + } standard technique is to use random bytes, not a fixed address like this, when MAC address is otherwise unavailable. consider what happens when two MACs are present. + for (i = 0; i < NODE_ADDRESS_SIZE; i++) + hw->mac_addr[i] = hw->perm_mac_addr[i]; + return ATL1_SUCCESS; +} + +/** + * Hashes an address to determine its location in the multicast table + * hw - Struct containing variables accessed by shared code + * mc_addr - the multicast address to hash + * + * atl1_hash_mc_addr + * purpose + * set hash value for a multicast address + * hash calcu processing : + * 1. calcu 32bit CRC for multicast address + * 2. reverse crc with MSB to LSB + **/ +u32 atl1_hash_mc_addr(struct atl1_hw *
Re: [2.6 patch] make hdlc_setup() static again
Adrian Bunk wrote: hdlc_setup was exported, but this export was never used. If a driver using it actually shows up it can still be exported again. Signed-off-by: Adrian Bunk <[EMAIL PROTECTED]> Acked-by: Krzysztof Halasa <[EMAIL PROTECTED]> applied - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 1/10] cxgb3 - main header files
Divy Le Ray wrote: Jeff Garzik wrote: Divy Le Ray wrote: From: Divy Le Ray <[EMAIL PROTECTED]> This patch implements the main header files of the Chelsio T3 network driver. Signed-off-by: Divy Le Ray <[EMAIL PROTECTED]> Once you think it's ready, email me a URL to a single patch that adds the driver to the latest linux-2.6.git kernel. Include in the email a description of the driver and signed-off-by line, which will get directly included in the git changelog. Adding new drivers is a bit special, because we want to merge it as a single changeset, but that would create a patch too large to review on the common kernel mailing lists. Jeff, You can grab the monolithic patch at this URL: http://service.chelsio.com/kernel.org/cxgb3.patch.bz2 applied to netdev-2.6.git#upstream I'm really counting on Chelsio to actively maintain this driver, unlike the abandonware you guys first submitted. Jeff - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC: 2.6 patch] remove the broken OAKNET driver
Adrian Bunk wrote: The OAKNET driver: - has been marked as BROKEN for more than two years and - is still marked as BROKEN. Drivers that had been marked as BROKEN for such a long time seem to be unlikely to be revived in the forseeable future. But if anyone wants to ever revive this driver, the code is still present in the older kernel releases. Signed-off-by: Adrian Bunk <[EMAIL PROTECTED]> applied - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
2.6.20-rc5: known regressions with patches (v2)
This email lists some known regressions in 2.6.20-rc5 compared to 2.6.19 with patches available If you find your name in the Cc header, you are either submitter of one of the bugs, maintainer of an affectected subsystem or driver, a patch of you caused a breakage or I'm considering you in any other way possibly involved with one or more of these issues. Due to the huge amount of recipients, please trim the Cc when answering. Subject: does not pickup ipv6 addresses References : http://bugzilla.kernel.org/show_bug.cgi?id=7817 http://lkml.org/lkml/2007/1/14/146 Submitter : Michael Gernoth <[EMAIL PROTECTED]> Daniel Drake <[EMAIL PROTECTED]> Caused-By : David L Stevens <[EMAIL PROTECTED]> commit 30c4cf577fb5b68c16e5750d6bdbd7072e42b279 Handled-By : YOSHIFUJI Hideaki <[EMAIL PROTECTED]> Patch : http://bugzilla.kernel.org/show_bug.cgi?id=7817 Status : patch available Subject: ACPI: fix cpufreq regression References : http://lkml.org/lkml/2007/1/16/120 Submitter : Ingo Molnar <[EMAIL PROTECTED]> Caused-By : Dave Jones <[EMAIL PROTECTED]> commit 0916bd3ebb7cefdd0f432e8491abe24f4b5a101e Handled-By : Ingo Molnar <[EMAIL PROTECTED]> Patch : http://lkml.org/lkml/2007/1/16/120 Status : patch available Subject: CONFIG_JFFS2_FS_DEBUG=2 compile error References : http://lkml.org/lkml/2007/1/12/161 Submitter : Russell King <[EMAIL PROTECTED]> Caused-By : Al Viro <[EMAIL PROTECTED]> commit 914e26379decf1fd984b22e51fd2e4209b7a7f1b Handled-By : David Woodhouse <[EMAIL PROTECTED]> Status : patch available Subject: WARNING: "profile_hits" [drivers/kvm/kvm-intel.ko] undefined! References : http://lkml.org/lkml/2007/1/12/16 Submitter : Miles Lane <[EMAIL PROTECTED]> Caused-By : Ingo Molnar <[EMAIL PROTECTED]> commit 07031e14c1127fc7e1a5b98dfcc59f434e025104 Handled-By : Andrew Morton <[EMAIL PROTECTED]> Patch : http://lkml.org/lkml/2007/1/12/18 Status : patch available Subject: KVM: guest crash References : http://lkml.org/lkml/2007/1/8/163 Submitter : Roland Dreier <[EMAIL PROTECTED]> Handled-By : Avi Kivity <[EMAIL PROTECTED]> Patch : http://lkml.org/lkml/2007/1/9/280 Status : patch available Subject: compile error: USB_HID must depend on INPUT References : http://lkml.org/lkml/2007/1/12/157 Submitter : Russell King <[EMAIL PROTECTED]> Handled-By : Russell King <[EMAIL PROTECTED]> Patch : http://lkml.org/lkml/2007/1/12/177 Status : patch available - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: 2.6.20-rc5: known regressions with patches (v2)
On Thu, 2007-01-18 at 20:59 +0100, Adrian Bunk wrote: > Subject: CONFIG_JFFS2_FS_DEBUG=2 compile error > References : http://lkml.org/lkml/2007/1/12/161 > Submitter : Russell King <[EMAIL PROTECTED]> > Caused-By : Al Viro <[EMAIL PROTECTED]> > commit 914e26379decf1fd984b22e51fd2e4209b7a7f1b > Handled-By : David Woodhouse <[EMAIL PROTECTED]> > Status : patch available Linus, please pull from git://git.infradead.org/mtd-2.6.git This fixes the above bug along with a few others. It does also contain a small amount of new code which has been waiting for a while (including the driver for the CAFÉ NAND controller which we use on OLPC.). My apologies for missing the merge window and first asking you to pull this a few hours after 2.6.20-rc1 was cut; I'd been waiting for the bitrev stuff to land, and had waited too long. Adrian Bunk (3): [MTD] SSFDC must depend on BLOCK [MTD] [NAND] rtc_from4.c: use lib/bitrev.c [MTD] make drivers/mtd/cmdlinepart.c:mtdpart_setup() static Adrian Hunter (2): [MTD] OneNAND: Implement read-while-load [MTD] OneNAND: Handle DDP chip boundary during read-while-load Akinobu Mita (1): [JFFS2] Use rb_first() and rb_last() cleanup Alan Cox (1): [MTD] MAPS: esb2rom: use hotplug safe interfaces Alexey Dobriyan (1): [MTD] JEDEC probe: fix comment typo (devic) Amit Choudhary (1): [JFFS2] Fix error-path leak in summary scan Andrew Morton (1): [MTD] Tidy bitrev usage in rtc_from4.c Andrew Victor (2): [MTD] NAND: AT91 NAND driver [MTD] NAND: Support for 16-bit bus-width on AT91. Artem Bityutskiy (10): [MTD] core: trivial comments fix [MTD] NAND: nandsim: support subpage write [MTD] increase MAX_MTD_DEVICES [MTD] add get_mtd_device_nm() function [MTD] add get and put methods [MTD] return error code from get_mtd_device() [MTD] nandsim: bugfix in page addressing [JFFS2] add cond_resched() when garbage collecting deletion dirent [JFFS2] Reschedule in loops [MTD] OneNAND: release CPU in cycles Burman Yan (1): [MTD] replace kmalloc+memset with kzalloc Dave Olsen (1): [MTD] [MAPS] Support for BIOS flash chips on the nvidia ck804 southbridge David Anders (1): [MTD] NOR: leave Intel chips in read-array mode on suspend David Woodhouse (29): [MTD NAND] Initial import of CAFÉ NAND driver. [MTD NAND] OLPC CAFÉ driver update Merge branch 'master' of git://git.kernel.org/.../torvalds/linux-2.6 [MTD] NAND: Combined oob buffer so it's contiguous with data [MTD] NAND: Correct setting of chip->oob_poi OOB buffer Merge git://git.infradead.org/~dwmw2/cafe-2.6 [MTD] NAND: Add hardware ECC correction support to CAFÉ NAND driver [MTD] NAND: CAFÉ NAND driver cleanup, fix ECC on reading empty flash [MTD] NAND: Disable ECC checking on CAFÉ since it's broken for now [MTD] NAND: Café ECC -- remove spurious BUG_ON() in err_pos() [MTD] NAND: Reset Café controller before initialising. [MTD] CAFÉ NAND: Add 'slowtiming' parameter, default usedma and checkecc on [MTD] NAND: Add ECC debugging for CAFÉ [MTD] NAND: Remove empty block ECC workaround [MTD] NAND: Fix timing calculation in CAFÉ debugging message [MTD] NAND: Use register #defines throughout CAFÉ driver, not numbers [MTD] NAND: Add register debugging spew option to CAFÉ driver [MTD] NAND: Fix ECC settings in CAFÉ controller driver. Merge git://git.infradead.org/~dwmw2/cafe-2.6 Merge git://git.infradead.org/~kmpark/onenand-mtd-2.6 [MTD] [NAND] Update CAFÉ driver interrupt handler prototype [MTD] Use EXPORT_SYMBOL_GPL() for exported symbols. [MTD] Remove trailing whitespace Merge branch 'master' of git://git.kernel.org/.../torvalds/linux-2.6 [MTD] Fix SSFDC build for variable blocksize. [MTD] Fix ssfdc blksize typo Merge branch 'master' of git://git.infradead.org/~kmpark/onenand-mtd-2.6 [JFFS2] debug.h: include for current->pid Merge branch 'master' of git://git.kernel.org/.../torvalds/linux-2.6 Haavard Skinnemoen (1): [MTD] bugfix: DataFlash is not bit writable Jeff Garzik (1): [JFFS2] kill warning RE debug-only variables Josh Boyer (1): [MTD] add MTD_BLKDEVS Kconfig option Kyungmin Park (9): MTD: OneNAND: interrupt based wait support [MTD] OneNAND: lock support [MTD] OneNAND: Single bit error detection [MTD] OneNAND: fix oob handling in recent oob patch [JFFS2] use the ref_offset macro [MTD] OneNAND: fix onenand_wait bug [MTD] OneNAND: add subpage write support [MTD] OneNAND: fix onenand_wait bug in read ecc error [MTD] OneNAND: return ecc error code only when 2-bit ecc occurs Lew Glendenning (1): [MTD] MAPS: Support for BIOS flash chips on Intel ESB2 southbridge Mariusz Kozlowski (1): [MTD] [NAND] Compile fix in rfc_from4.c Qi
Re: [take33 10/10] kevent: Kevent based AIO (aio_sendfile()/aio_sendfile_path()).
On Wed, Jan 17, 2007 at 05:39:51PM +0300, Evgeniy Polyakov wrote: > On Wed, Jan 17, 2007 at 07:21:42PM +0530, Suparna Bhattacharya ([EMAIL > PROTECTED]) wrote: > > > > Since you are implementing new APIs here, have you considered doing an > > aio_sendfilev to be able to send a header with the data ? > > It is doable, but why people do not like corking? > With Linux less than microsecond syscall overhead it is better and more > flexible solution, doesn't it? That is what I used to think as well. However ... The problem as I understand it now is not about bunching data together, but of ensuring some sort of atomicity between the header and the data, when there can be multiple outstanding aio requests on the same socket - i.e ensuring strict ordering without other data coming in between, when data to be sent is not already in cache, and in the meantime another sendfile or aio write requests comes in for the same socket. Without having to lock the socket when reading data from disk. There are alternate ways to address this, aio_sendfilev is one of the options I have heard people requesting. Regards Suparna > > I'm not saying - 'no, there will not be any *v variants', just getting > more info. > > > Regards > > Suparna > > -- > Evgeniy Polyakov -- Suparna Bhattacharya ([EMAIL PROTECTED]) Linux Technology Center IBM Software Lab, India - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH 0/12] L2 network namespace (v3)
YOSHIFUJI Hideaki / 吉藤英明 <[EMAIL PROTECTED]> writes: > In article <[EMAIL PROTECTED]> (at Wed, 17 Jan 2007 18:51:14 > +0300), Dmitry Mishin <[EMAIL PROTECTED]> says: > >> === >> L2 network namespaces >> >> The most straightforward concept of network virtualization is complete >> separation of namespaces, covering device list, routing tables, netfilter >> tables, socket hashes, and everything else. >> >> On input path, each packet is tagged with namespace right from the >> place where it appears from a device, and is processed by each layer >> in the context of this namespace. >> Non-root namespaces communicate with the outside world in two ways: by >> owning hardware devices, or receiving packets forwarded them by their parent >> namespace via pass-through device. > > Can you handle multicast / broadcast and IPv6, which are very important? The basic idea here is very simple. Each network namespace appears to user space as a separate network stack, with it's own set of routing tables etc. All sockets and all network devices (the sources of packets) belong to exactly one network namespace. >From the socket or the network device a packet enters the network stack you can infer the network namespace that it will be processed in. Each network namespace should get it own complement of the data structures necessary to process packets, and everything should work. Talking between namespaces is accomplished either through an external network, or through a special pseudo network device. The simplest to implement is two network devices where all packets transmitted on one are received on the other. Then by placing one network device in one namespace and the other in another interface it looks like two machines connected by a cross over cable. Once you have that in a one namespace you can connect other namespaces with the existing ethernet bridging or by configuring one of the namespaces as a router and routing traffic between them. Supporting IPv6 is roughly as difficult as supporting IPv4. What needs to happen to convert code is all variables either need a per network namespace instance or the data structures needs to be modified to have a network namespace tag. For hash tables which are hard to allocate dynamically tagging is the preferred conversion method, for anything that is small enough duplication is preferred as it allows the existing logic to be kept. In the fast path the impact of all of the conversions should be very light, to non-existent. In network stack initialization and cleanup there is work todo because you are initializing and cleanup variables more often then at module insertion and removal. So my expectation is that once we get a framework established and merged to allow network namespaces eventually the entire network stack will be converted. Not just ipv4 and ipv6 but decnet, ipx, iptables, fair scheduling, ethernet bridging and all of the other weird and twisty bits of the linux network stack. The primary practical hurdle is there is a lot of networking code in the kernel. I think I know a path by which we can incrementally merge support for network namespaces without breaking anything. More to come on this when I finish up my demonstration patchset in a week or so that is complete enough to show what I am talking about. I hope this helps but the concept into perspective. As for Dmitry's patchset in particular it currently does not support IPv6 and I don't know where it is with respect to the broadcast and multicast but I don't see any immediate problems that would preclude those from working. But any incompleteness is exactly that incompleteness and an implementation problem not a fundamental design issue. Eric - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
[PATCH 2.6.20 1/5] s2io: Making LRO and UFO as module loadable parameter.
This patch adds two load parameters napi and ufo. Previously NAPI was compilation option with these changes wan enable disable NAPI using load parameter. Also we are introducing ufo load parameter to enable/disable ufo feature Signed-off-by: Sivakumar Subramani <[EMAIL PROTECTED]> --- diff -urpN orig/drivers/net/s2io.c patch1/drivers/net/s2io.c --- orig/drivers/net/s2io.c 2006-12-21 10:06:58.0 +0530 +++ patch1/drivers/net/s2io.c 2007-01-08 11:56:23.0 +0530 @@ -401,9 +401,10 @@ S2IO_PARM_INT(lro, 0); * aggregation happens until we hit max IP pkt size(64K) */ S2IO_PARM_INT(lro_max_pkts, 0x); -#ifndef CONFIG_S2IO_NAPI S2IO_PARM_INT(indicate_max_pkts, 0); -#endif + +S2IO_PARM_INT(napi, 1); +S2IO_PARM_INT(ufo, 0); static unsigned int tx_fifo_len[MAX_TX_FIFOS] = {DEFAULT_FIFO_0_LEN, [1 ...(MAX_TX_FIFOS - 1)] = DEFAULT_FIFO_1_7_LEN}; @@ -2275,9 +2276,7 @@ static int fill_rx_buffers(struct s2io_n struct config_param *config; u64 tmp; buffAdd_t *ba; -#ifndef CONFIG_S2IO_NAPI unsigned long flags; -#endif RxD_t *first_rxdp = NULL; mac_control = &nic->mac_control; @@ -2321,12 +2320,15 @@ static int fill_rx_buffers(struct s2io_n DBG_PRINT(INTR_DBG, "%s: Next block at: %p\n", dev->name, rxdp); } -#ifndef CONFIG_S2IO_NAPI - spin_lock_irqsave(&nic->put_lock, flags); - mac_control->rings[ring_no].put_pos = - (block_no * (rxd_count[nic->rxd_mode] + 1)) + off; - spin_unlock_irqrestore(&nic->put_lock, flags); -#endif + if(!napi) { + spin_lock_irqsave(&nic->put_lock, flags); + mac_control->rings[ring_no].put_pos = + (block_no * (rxd_count[nic->rxd_mode] + 1)) + off; + spin_unlock_irqrestore(&nic->put_lock, flags); + } else { + mac_control->rings[ring_no].put_pos = + (block_no * (rxd_count[nic->rxd_mode] + 1)) + off; + } if ((rxdp->Control_1 & RXD_OWN_XENA) && ((nic->rxd_mode >= RXD_MODE_3A) && (rxdp->Control_2 & BIT(0 { @@ -2569,7 +2571,7 @@ static void free_rx_buffers(struct s2io_ * 0 on success and 1 if there are No Rx packets to be processed. */ -#if defined(CONFIG_S2IO_NAPI) +#if defined(HAVE_NETDEV_POLL) static int s2io_poll(struct net_device *dev, int *budget) { nic_t *nic = dev->priv; @@ -2708,9 +2710,7 @@ static void rx_intr_handler(ring_info_t rx_curr_get_info_t get_info, put_info; RxD_t *rxdp; struct sk_buff *skb; -#ifndef CONFIG_S2IO_NAPI int pkt_cnt = 0; -#endif int i; spin_lock(&nic->rx_lock); @@ -2726,16 +2726,16 @@ static void rx_intr_handler(ring_info_t put_info = ring_data->rx_curr_put_info; put_block = put_info.block_index; rxdp = ring_data->rx_blocks[get_block].rxds[get_info.offset].virt_addr; -#ifndef CONFIG_S2IO_NAPI - spin_lock(&nic->put_lock); - put_offset = ring_data->put_pos; - spin_unlock(&nic->put_lock); -#else - put_offset = (put_block * (rxd_count[nic->rxd_mode] + 1)) + - put_info.offset; -#endif + if (!napi) { + spin_lock(&nic->put_lock); + put_offset = ring_data->put_pos; + spin_unlock(&nic->put_lock); + } else + put_offset = ring_data->put_pos; + while (RXD_IS_UP2DT(rxdp)) { - /* If your are next to put index then it's FIFO full condition */ + /* If your are next to put index then it's + FIFO full condition */ if ((get_block == put_block) && (get_info.offset + 1) == put_info.offset) { DBG_PRINT(INTR_DBG, "%s: Ring Full\n",dev->name); @@ -2793,15 +2793,12 @@ static void rx_intr_handler(ring_info_t rxdp = ring_data->rx_blocks[get_block].block_virt_addr; } -#ifdef CONFIG_S2IO_NAPI nic->pkts_to_process -= 1; - if (!nic->pkts_to_process) + if ((napi) && (!nic->pkts_to_process)) break; -#else pkt_cnt++; if ((indicate_max_pkts) && (pkt_cnt > indicate_max_pkts)) break; -#endif } if (nic->lro) { /* Clear all LRO sessions before exiting */ @@ -4196,26 +4193,26 @@ static irqreturn_t s2io_isr(int irq, voi org_mask = readq(&bar0->general_int_mask); writeq(val64, &bar0->general_int_mask); -#ifdef CONFIG_S2IO_NAPI - if (reason & GEN_INTR_RXTRAFFIC) { - if (netif_rx_schedule_prep(dev)) { - writeq(val64, &bar0->rx_traffic_mask); - __netif_rx_schedule(dev); +
[PATCH 2.6.20 2/5] S2IO: Fixes for reset and link handling.
1. Fix for reset and link handling. 2. Allow for promiscuos mode and multicast state be maintained through ifconfig up and down. 3. Support to print adapter serial number. Signed-off-by: Sivakumar Subramani <[EMAIL PROTECTED]> --- diff -urpN patch1/drivers/net/s2io.c patch2/drivers/net/s2io.c --- patch1/drivers/net/s2io.c 2007-01-08 11:56:23.0 +0530 +++ patch2/drivers/net/s2io.c 2007-01-08 17:51:20.0 +0530 @@ -1416,7 +1416,7 @@ static int init_nic(struct s2io_nic *nic val64 = TTI_DATA2_MEM_TX_UFC_A(0x10) | TTI_DATA2_MEM_TX_UFC_B(0x20) | - TTI_DATA2_MEM_TX_UFC_C(0x70) | TTI_DATA2_MEM_TX_UFC_D(0x80); + TTI_DATA2_MEM_TX_UFC_C(0x40) | TTI_DATA2_MEM_TX_UFC_D(0x80); writeq(val64, &bar0->tti_data2_mem); val64 = TTI_CMD_MEM_WE | TTI_CMD_MEM_STROBE_NEW_CMD; @@ -1612,7 +1612,8 @@ static int init_nic(struct s2io_nic *nic * that does not start on an ADB to reduce disconnects. */ if (nic->device_type == XFRAME_II_DEVICE) { - val64 = EXT_REQ_EN | MISC_LINK_STABILITY_PRD(3); + val64 = FAULT_BEHAVIOUR | EXT_REQ_EN | + MISC_LINK_STABILITY_PRD(3); writeq(val64, &bar0->misc_control); val64 = readq(&bar0->pic_control2); val64 &= ~(BIT(13)|BIT(14)|BIT(15)); @@ -1879,41 +1880,36 @@ static void en_dis_able_nic_intrs(struct } } -static int check_prc_pcc_state(u64 val64, int flag, int rev_id, int herc) +/** + * verify_pcc_quiescent- Checks for PCC quiescent state + * Return: 1 If PCC is quiescence + * 0 If PCC is not quiescence + */ +static int verify_pcc_quiescent(nic_t *sp, int flag) { - int ret = 0; + int ret = 0, herc; + XENA_dev_config_t __iomem *bar0 = sp->bar0; + u64 val64 = readq(&bar0->adapter_status); + + herc = (sp->device_type == XFRAME_II_DEVICE); if (flag == FALSE) { - if ((!herc && (rev_id >= 4)) || herc) { - if (!(val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) && - ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) == -ADAPTER_STATUS_RC_PRC_QUIESCENT)) { + if ((!herc && (get_xena_rev_id(sp->pdev) >= 4)) || herc) { + if (!(val64 & ADAPTER_STATUS_RMAC_PCC_IDLE)) ret = 1; - } - }else { - if (!(val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) && - ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) == -ADAPTER_STATUS_RC_PRC_QUIESCENT)) { + } else { + if (!(val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE)) ret = 1; - } } } else { - if ((!herc && (rev_id >= 4)) || herc) { + if ((!herc && (get_xena_rev_id(sp->pdev) >= 4)) || herc) { if (((val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) == -ADAPTER_STATUS_RMAC_PCC_IDLE) && - (!(val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) || -((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) == - ADAPTER_STATUS_RC_PRC_QUIESCENT))) { +ADAPTER_STATUS_RMAC_PCC_IDLE)) ret = 1; - } } else { if (((val64 & ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) == -ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE) && - (!(val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) || -((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) == - ADAPTER_STATUS_RC_PRC_QUIESCENT))) { +ADAPTER_STATUS_RMAC_PCC_FOUR_IDLE)) ret = 1; - } } } @@ -1921,9 +1917,6 @@ static int check_prc_pcc_state(u64 val64 } /** * verify_xena_quiescence - Checks whether the H/W is ready - * @val64 : Value read from adapter status register. - * @flag : indicates if the adapter enable bit was ever written once - * before. * Description: Returns whether the H/W is ready to go or not. Depending * on whether adapter enable bit was written or not the comparison * differs and the calling function passes the input argument flag to @@ -1932,24 +1925,63 @@ static int check_prc_pcc_state(u64 val64 * 0 If Xena is not quiescence */ -static int verify_xena_quiescence(nic_t *sp, u64 val64, int flag) +static int verify_xena_quiescence(nic_t *sp) { - int ret = 0, herc; - u64 tmp64 = ~((u64) val64); - int rev_id = get_xena_rev_id(sp->pdev); + int mode; + XENA_dev_config_t __iomem *bar0 = sp->bar0; + u64 val64 = readq(&bar0->adapter_status); + mode
[PATCH 2.6.20 3/5] s2io: Fixes in updating skb->truesize and code cleanup.
1. Fix for updating skb->truesize properly. 2. Disable NAPI only if more than one ring configured in case of MSI/MSI-X interrupts. Previously we were disabling NAPI irrespective of number of rings when MSI/MSI-X interrupts were used. 3. Code cleanup. Signed-off-by: Sivakumar Subramani <[EMAIL PROTECTED]> --- diff -urpN patch2/drivers/net/s2io.c patch3/drivers/net/s2io.c --- patch2/drivers/net/s2io.c 2007-01-08 17:51:20.0 +0530 +++ patch3/drivers/net/s2io.c 2007-01-08 23:38:48.0 +0530 @@ -459,7 +459,7 @@ static int init_shared_mem(struct s2io_n void *tmp_v_addr, *tmp_v_addr_next; dma_addr_t tmp_p_addr, tmp_p_addr_next; RxD_block_t *pre_rxd_blk = NULL; - int i, j, blk_cnt, rx_sz, tx_sz; + int i, j, blk_cnt; int lst_size, lst_per_page; struct net_device *dev = nic->dev; unsigned long tmp; @@ -484,7 +484,6 @@ static int init_shared_mem(struct s2io_n } lst_size = (sizeof(TxD_t) * config->max_txds); - tx_sz = lst_size * size; lst_per_page = PAGE_SIZE / lst_size; for (i = 0; i < config->tx_fifo_num; i++) { @@ -584,7 +583,6 @@ static int init_shared_mem(struct s2io_n size = (size * (sizeof(RxD1_t))); else size = (size * (sizeof(RxD3_t))); - rx_sz = size; for (i = 0; i < config->rx_ring_num; i++) { mac_control->rings[i].rx_curr_get_info.block_index = 0; @@ -625,6 +623,8 @@ static int init_shared_mem(struct s2io_n rx_blocks->rxds = kmalloc(sizeof(rxd_info_t)* rxd_count[nic->rxd_mode], GFP_KERNEL); + if (!rx_blocks->rxds) + return -ENOMEM; for (l=0; lrxd_mode];l++) { rx_blocks->rxds[l].virt_addr = rx_blocks->block_virt_addr + @@ -2260,6 +2260,7 @@ static int fill_rxd_3buf(nic_t *nic, RxD return -ENOMEM ; } frag_list = skb_shinfo(skb)->frag_list; + skb->truesize += frag_list->truesize; //updating skb->truesize frag_list->next = NULL; tmp = (void *)ALIGN((long)frag_list->data, ALIGN_SIZE + 1); frag_list->data = tmp; @@ -3186,6 +3187,8 @@ static void alarm_intr_handler(struct s2 register u64 val64 = 0, err_reg = 0; u64 cnt; int i; + if (atomic_read(&nic->card_state) == CARD_DOWN) + return; nic->mac_control.stats_info->sw_stat.ring_full_cnt = 0; /* Handling the XPAK counters update */ if(nic->mac_control.stats_info->xpak_stat.xpak_timer_count < 72000) { @@ -6581,7 +6584,6 @@ static int rx_osm_handler(ring_info_t *r skb_put(skb, buf1_len); skb->len += buf2_len; skb->data_len += buf2_len; - skb->truesize += buf2_len; skb_put(skb_shinfo(skb)->frag_list, buf2_len); sp->stats.rx_bytes += buf1_len; @@ -6803,6 +6805,8 @@ static int s2io_verify_parm(struct pci_d "Defaulting to INTA\n"); *dev_intr_type = INTA; } + if ( (rx_ring_num > 1) && (*dev_intr_type != INTA) ) + napi = 0; if (rx_ring_mode > 3) { DBG_PRINT(ERR_DBG, "s2io: Requested ring mode not supported\n"); DBG_PRINT(ERR_DBG, "s2io: Defaulting to 3-buffer mode\n"); @@ -6999,7 +7003,7 @@ s2io_init_nic(struct pci_dev *pdev, cons goto mem_alloc_failed; } - sp->bar0 = ioremap(pci_resource_start(pdev, 0), + sp->bar0 = (caddr_t) ioremap(pci_resource_start(pdev, 0), pci_resource_len(pdev, 0)); if (!sp->bar0) { DBG_PRINT(ERR_DBG, "%s: Neterion: cannot remap io mem1\n", @@ -7008,7 +7012,7 @@ s2io_init_nic(struct pci_dev *pdev, cons goto bar0_remap_failed; } - sp->bar1 = ioremap(pci_resource_start(pdev, 2), + sp->bar1 = (caddr_t) ioremap(pci_resource_start(pdev, 2), pci_resource_len(pdev, 2)); if (!sp->bar1) { DBG_PRINT(ERR_DBG, "%s: Neterion: cannot remap io mem2\n", @@ -7324,7 +7328,7 @@ int __init s2io_starter(void) * Description: This function is the cleanup routine for the driver. It unregist * ers the driver. */ -static void s2io_closer(void) +static __exit void s2io_closer(void) { pci_unregister_driver(&s2io_driver); DBG_PRINT(INIT_DBG, "cleanup done\n"); @@ -7645,6 +7649,7 @@ static void lro_append_pkt(nic_t *sp, lr lro->last_frag->next = skb; else skb_shinfo(first)->frag_list = skb; + first->truesize += skb->truesize;//updating skb->truesize lro->l
[PATCH 2.6.20 4/5] s2io: Removed enabling of some of the unused interrupts.
Removed unused code in en_dis_able_nic_intrs(), TX_DMA_INTR, RX_DMA_INTR, TX_XGXS_INTR, MC_INTR Signed-off-by: Sivakumar Subramani <[EMAIL PROTECTED]> --- diff -urpN patch3/drivers/net/s2io.c patch4/drivers/net/s2io.c --- patch3/drivers/net/s2io.c 2007-01-08 23:38:48.0 +0530 +++ patch4/drivers/net/s2io.c 2007-01-09 01:12:42.0 +0530 @@ -1659,7 +1659,7 @@ static void en_dis_able_nic_intrs(struct /* PIC Interrupts */ if ((mask & (TX_PIC_INTR | RX_PIC_INTR))) { /* Enable PIC Intrs in the general intr mask register */ - val64 = TXPIC_INT_M | PIC_RX_INT_M; + val64 = TXPIC_INT_M; if (flag == ENABLE_INTRS) { temp64 = readq(&bar0->general_int_mask); temp64 &= ~((u64) val64); @@ -1697,70 +1697,6 @@ static void en_dis_able_nic_intrs(struct } } - /* DMA Interrupts */ - /* Enabling/Disabling Tx DMA interrupts */ - if (mask & TX_DMA_INTR) { - /* Enable TxDMA Intrs in the general intr mask register */ - val64 = TXDMA_INT_M; - if (flag == ENABLE_INTRS) { - temp64 = readq(&bar0->general_int_mask); - temp64 &= ~((u64) val64); - writeq(temp64, &bar0->general_int_mask); - /* -* Keep all interrupts other than PFC interrupt -* and PCC interrupt disabled in DMA level. -*/ - val64 = DISABLE_ALL_INTRS & ~(TXDMA_PFC_INT_M | - TXDMA_PCC_INT_M); - writeq(val64, &bar0->txdma_int_mask); - /* -* Enable only the MISC error 1 interrupt in PFC block -*/ - val64 = DISABLE_ALL_INTRS & (~PFC_MISC_ERR_1); - writeq(val64, &bar0->pfc_err_mask); - /* -* Enable only the FB_ECC error interrupt in PCC block -*/ - val64 = DISABLE_ALL_INTRS & (~PCC_FB_ECC_ERR); - writeq(val64, &bar0->pcc_err_mask); - } else if (flag == DISABLE_INTRS) { - /* -* Disable TxDMA Intrs in the general intr mask -* register -*/ - writeq(DISABLE_ALL_INTRS, &bar0->txdma_int_mask); - writeq(DISABLE_ALL_INTRS, &bar0->pfc_err_mask); - temp64 = readq(&bar0->general_int_mask); - val64 |= temp64; - writeq(val64, &bar0->general_int_mask); - } - } - - /* Enabling/Disabling Rx DMA interrupts */ - if (mask & RX_DMA_INTR) { - /* Enable RxDMA Intrs in the general intr mask register */ - val64 = RXDMA_INT_M; - if (flag == ENABLE_INTRS) { - temp64 = readq(&bar0->general_int_mask); - temp64 &= ~((u64) val64); - writeq(temp64, &bar0->general_int_mask); - /* -* All RxDMA block interrupts are disabled for now -* TODO -*/ - writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask); - } else if (flag == DISABLE_INTRS) { - /* -* Disable RxDMA Intrs in the general intr mask -* register -*/ - writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask); - temp64 = readq(&bar0->general_int_mask); - val64 |= temp64; - writeq(val64, &bar0->general_int_mask); - } - } - /* MAC Interrupts */ /* Enabling/Disabling MAC interrupts */ if (mask & (TX_MAC_INTR | RX_MAC_INTR)) { @@ -1787,53 +1723,6 @@ static void en_dis_able_nic_intrs(struct } } - /* XGXS Interrupts */ - if (mask & (TX_XGXS_INTR | RX_XGXS_INTR)) { - val64 = TXXGXS_INT_M | RXXGXS_INT_M; - if (flag == ENABLE_INTRS) { - temp64 = readq(&bar0->general_int_mask); - temp64 &= ~((u64) val64); - writeq(temp64, &bar0->general_int_mask); - /* -* All XGXS block error interrupts are disabled for now -* TODO -*/ - writeq(DISABLE_ALL_INTRS, &bar0->xgxs_int_mask); - } else if (flag == DISABLE_INTRS) { - /* -* Disable MC Intrs in the general intr mask register -
[PATCH] bcm43xx: Check error returns in initialization routines
A number of the calls in the initialization routines fail to check the returned value for errors. This patch adds the necessary checks and logs any errors found when appropriate. Signed-off-by: Larry Finger <[EMAIL PROTECTED]> --- To be applied to wireless-2.6. Index: wireless-2.6/drivers/net/wireless/bcm43xx/bcm43xx_main.c === --- wireless-2.6.orig/drivers/net/wireless/bcm43xx/bcm43xx_main.c +++ wireless-2.6/drivers/net/wireless/bcm43xx/bcm43xx_main.c @@ -2980,8 +2980,10 @@ static int bcm43xx_chipset_attach(struct err = bcm43xx_pctl_set_crystal(bcm, 1); if (err) goto out; - bcm43xx_pci_read_config16(bcm, PCI_STATUS, &pci_status); - bcm43xx_pci_write_config16(bcm, PCI_STATUS, pci_status & ~PCI_STATUS_SIG_TARGET_ABORT); + err = bcm43xx_pci_read_config16(bcm, PCI_STATUS, &pci_status); + if (err) + goto out; + err = bcm43xx_pci_write_config16(bcm, PCI_STATUS, pci_status & ~PCI_STATUS_SIG_TARGET_ABORT); out: return err; @@ -3793,12 +3795,18 @@ static int bcm43xx_attach_board(struct b } net_dev->base_addr = (unsigned long)bcm->mmio_addr; - bcm43xx_pci_read_config16(bcm, PCI_SUBSYSTEM_VENDOR_ID, + err = bcm43xx_pci_read_config16(bcm, PCI_SUBSYSTEM_VENDOR_ID, &bcm->board_vendor); - bcm43xx_pci_read_config16(bcm, PCI_SUBSYSTEM_ID, + if (err) + goto err_iounmap; + err = bcm43xx_pci_read_config16(bcm, PCI_SUBSYSTEM_ID, &bcm->board_type); - bcm43xx_pci_read_config16(bcm, PCI_REVISION_ID, + if (err) + goto err_iounmap; + err = bcm43xx_pci_read_config16(bcm, PCI_REVISION_ID, &bcm->board_revision); + if (err) + goto err_iounmap; err = bcm43xx_chipset_attach(bcm); if (err) @@ -3889,6 +3897,7 @@ err_pci_release: pci_release_regions(pci_dev); err_pci_disable: pci_disable_device(pci_dev); + printk(KERN_ERR PFX "Unable to attach board\n"); goto out; } --- - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: Can someone please try...
Hello, Michael! I did more testing, and the results are following. It looks like the oopses and panics on i386 were triggered by 4k stacks. x86_64 doesn't have this option. Now that I enabled other debug options on both platforms. but not 4k stacks, I'm seeing exactly the same problem on each platform. When run initially, wpa_supplicant connects with no problems (except very poor reception of the data packets, but it's another story). If interrupted and restarted, wpa_supplicant reconnects, but I'm getting messages like this (i386): Slab corruption: start=cfdaece0, len=1024 Redzone: 0x5a2cf071/0x5a2cf071. Last user: [](skb_release_data+0x7b/0x7f) 000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 050: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 Prev obj: start=cfdae8d4, len=1024 Redzone: 0x170fc2a5/0x170fc2a5. Last user: [](device_create+0x2c/0x98) 000: 01 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 010: ad 4e ad de ff ff ff ff ff ff ff ff 10 3a 6d c0 Next obj: start=cfdaf0ec, len=1024 Redzone: 0x170fc2a5/0x170fc2a5. Last user: [](expand_files+0x95/0x2c2) 000: 78 55 39 c7 78 55 39 c7 78 55 39 c7 88 da 52 df 010: d8 18 3b c7 00 00 00 00 00 00 00 00 00 00 00 00 and this (x86_64): Slab corruption: start=81000ec8a198, len=1024 Redzone: 0x5a2cf071/0x5a2cf071. Last user: [](skb_release_data+0x94/0x99) 000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 050: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 Next obj: start=81000ec8a5b0, len=1024 Redzone: 0x170fc2a5/0x170fc2a5. Last user: [](device_create+0x5f/0x110) 000: 01 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 I can restart wpa_supplicant again, and it would show similar messages. The first "Last user" is inevitably skb_release_data. I have no idea how to deal with it. I think I need a stack trace at the time when skb_release_data is called. This is a stack trace at the time when slab corruption is detected. It's actually incorrect closer to the top, perhaps from gcc optimizations for static functions. Slab corruption: start=8100066f81d8, len=1024 Call Trace: [] vsnprintf+0x338/0x5a8 [] check_poison_obj+0x69/0x1ae [] _request_firmware+0x8f/0x326 [] _request_firmware+0x8f/0x326 [] cache_alloc_debugcheck_after+0x32/0x1a2 [] _request_firmware+0x8f/0x326 [] kmem_cache_zalloc+0xaf/0xd8 [] _request_firmware+0x8f/0x326 [] :bcm43xx_d80211:bcm43xx_phy_init_tssi2dbm_table +0xf0/0x2ca [] request_firmware+0xe/0x10 [] :bcm43xx_d80211:bcm43xx_chip_init+0x96/0xaba [] kmem_cache_alloc+0xaf/0xbe [] :bcm43xx_d80211:bcm43xx_wireless_core_init +0x4de/0xa3d [] :bcm43xx_d80211:bcm43xx_add_interface+0x64/0xde [] ieee80211_open+0x1c7/0x2cc [] dev_open+0x36/0x76 [] dev_change_flags+0x5d/0x122 [] devinet_ioctl+0x259/0x5e8 [] inet_ioctl+0x71/0x8f [] sock_ioctl+0x1db/0x1fd [] do_ioctl+0x1b/0x50 [] vfs_ioctl+0x22a/0x23c [] trace_hardirqs_on+0x124/0x14e [] sys_ioctl+0x42/0x65 [] system_call+0x7e/0x83 Anyway, I could narrow down this message to the first kzalloc() call in fw_register_device(), file drivers/base/firmware_class.c. This only seems to confirm my suspicion that the actual corruption happened before this point. We are just hitting it when trying to allocate more memory. Help with debugging this problem will be appreciated. I've never hunted down such problems, especially in kernel space. -- Regards, Pavel Roskin - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html