linux/drivers/net/spider_net.c

2606 lines
70 KiB

[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 16:11:33 -05:00
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 16:11:33 -05:00
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 16:11:33 -05:00
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 16:11:33 -05:00
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 16:11:33 -05:00
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 16:11:33 -05:00
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 16:11:33 -05:00
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 16:11:33 -05:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 14:55:46 +01:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 14:55:46 +01:00
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 16:11:33 -05:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 16:41:36 -07:00
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 13:21:13 -05:00