JP6195316B2 - Information processing apparatus, information processing method, and program - Google Patents

Description

  The present invention relates to an information processing apparatus, an information processing method, and a program.

  For example, Patent Document 1 discloses a multi-core system that can determine a failure of a shared resource using an adjustment system, and control reset and access to the shared resource.

US Patent Application Publication No. 2010/0325495

  An object of the present invention is to provide an information processing apparatus capable of restarting a LAN card while suppressing the influence on other modules.

  An information processing apparatus according to the present invention includes: a transmission / reception unit that transmits / receives data via a LAN card; and a lower-level transmission / reception unit that accesses the LAN card and transmits / receives data according to control information from the transmission / reception unit A detection unit that detects an abnormality of the LAN card; and a return instruction unit that instructs a restart operation to the LAN card when an abnormality is detected by the detection unit. While the restart operation instructed by the return instruction unit is being executed, control information for prohibiting at least transmission processing is output to the lower-level transmission / reception unit.

  Preferably, the detection unit detects an abnormality of the LAN card based on a state of a transmission queue managed by the transmission / reception unit.

  Preferably, the return instruction unit is a single control plane core, the transmission / reception unit includes a plurality of data plane cores, and all the data plane cores at least transmit control information for prohibiting transmission processing. Output to the lower transmission unit.

  Preferably, the detection unit determines that the LAN card is abnormal when the transmission queue does not change for a predetermined period in any of the data plane cores.

  Preferably, the control plane core ignores a request from an upper layer while a restart operation of the LAN card is being performed, and terminates a LAN card termination process, an initialization process, and Control information for sequentially executing the start processing is output, and the restarting operation of the LAN card is realized.

  An information processing method according to the present invention includes a transmission / reception step for transmitting / receiving data via a LAN card, and a lower transmission / reception step for transmitting / receiving data in accordance with control information from the transmission / reception step. A detection step of detecting an abnormality of the LAN card, a return instruction step of instructing a restart operation to the LAN card when an abnormality of the LAN card is detected, and the restart of the LAN card While the operation is being executed, an output step for outputting at least control information for prohibiting the transmission process is provided for the lower transmission / reception step for performing the data transmission / reception process.

  A program according to the present invention includes a transmission / reception function for performing data transmission / reception processing via a LAN card, a lower-level transmission / reception function for accessing the LAN card and transmitting / receiving data according to control information from the transmission / reception function, A detection function for detecting an abnormality of the LAN card, a return instruction function for instructing a restart operation to the LAN card when an abnormality of the LAN card is detected, and the restart operation of the LAN card. While the program is being executed, the computer realizes at least an output function for outputting control information for prohibiting the transmission process for the lower-level transmission / reception function for performing the data transmission / reception process.

  The LAN card can be restarted while suppressing the influence on other modules.

1 is a diagram illustrating a hardware configuration of a network communication system 1 including a network device 10. FIG. 2 is a diagram illustrating a hardware configuration of a network device 10. FIG. 2 is a diagram illustrating a functional block of a network device 10. FIG. FIG. 4 is a diagram illustrating a detailed functional configuration of a LAN driver 360 (control plane core side) and a LAN driver 460 (data plane core side) in FIG. 3. (A) is a figure which illustrates the system configuration centering on the relationship between a control plane core and a data plane core, and the interface of a LAN card, (B) tried paying attention to a single interface It is a figure which illustrates the system configuration in the case. It is a figure which illustrates typically the packet transmission / reception process by the data plane core process part 400 of a normal state. It is a figure which illustrates typically the state of the data plane core process part 400 during a LAN card restart. It is a flowchart explaining the abnormality detection process (S10) by the side of a data plane core. It is a flowchart explaining the abnormality detection process (S20) by the side of a control plane core. It is a flowchart explaining the restart process (S30) of a LAN card.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a diagram illustrating a hardware configuration of a network communication system 1 including a network device 10.
As illustrated in FIG. 1, the network communication system 1 includes a plurality of network devices 10 and a network switch 20 that connects the network devices 10.
The network device 10 is a network device used by connecting to a network. The network device 10 is an example of an information processing device according to the present invention. The network device of this example is connected to the LAN via the LAN cable and the network switch 20 and has a bandwidth control function, a firewall function, etc., and is redundant with other network devices 10 to provide fault tolerance through a redundant configuration. It is increasing. Specifically, when the survival monitoring packet is transmitted and received between the operating network device 10A (Active) and the waiting network device 10B (Standby) and the predetermined time survival monitoring packet is not transmitted and received, The network device 10B becomes operational and takes over the role.

  For example, in the network communication system 1 of this example, a survival monitoring packet is transmitted between two network devices 10 at intervals of 1 second, and if the survival monitoring packet is transmitted for 3 seconds or more, an error is detected by duplex monitoring. , Active / Standby network device 10 is set to switch. In this case, if a transmission timeout is detected due to an abnormality in the LAN card and it takes 3 seconds or more to recover after reset recovery, the survival monitoring packet cannot be transmitted between the network devices 10 and can be recovered. Although it is an event, Active / Standby system switching occurs. Therefore, in this example, it is desirable to complete the transmission timeout detection to reset recovery within 3 seconds.

FIG. 2 is a diagram illustrating a hardware configuration of the network device 10.
As illustrated in FIG. 2, the network device 10 includes a CPU 100, a memory 102, a plurality of slots 104, a south bridge 106, a hard disk drive 108, a compact flash conversion adapter 110, and a DVD drive 112.
The CPU 100 is a central processing unit. The CPU 100 in this example is a multi-core processor including a plurality of cores.
The memory 102 is, for example, a semiconductor memory, is connected to the CPU 100, and functions as a main storage device.
The slot 104 is an expansion slot, and for example, a LAN card 120 or the like is inserted therein. A LAN card 120 is inserted into the slot 104 of this example, and the LAN card 120 of this example is provided with a plurality of interfaces (# 0 to # 3) via PCI switches. Here, the LAN card means a network card in a broad sense, and is a concept including not only a LAN card in a narrow sense for Ethernet (registered trademark) but also a network card such as a fiber channel card, a network adapter, and the like.
The south bridge 106 is an IC chip that connects, for example, a SATA bus. In this example, a hard disk drive 108, a compact flash conversion adapter 110, and a DVD drive 112 are connected to the south bridge 106.
The network apparatus 10 of this example employs a data plane development kit (Data Plane Development Kit) in the hardware configuration described above, thereby realizing high-speed packet processing.

Here, in the data plane development kit that is open source software, a transmission timeout detection process and a reset process when the transmission process is executed across a plurality of cores (a plurality of data plane cores) are not prepared. In other words, the LAN driver (PMD: Polling Mode Driver) employed in the data plane development kit is separated into two functions: a control plane core and a data plane core. The control plane core mainly controls the LAN card, and the data plane core uses a special CPU core that performs packet transmission / reception processing. Furthermore, in order to process packets at high speed, there are a plurality of data plane cores, and for each interface on one LAN card, a transmission / reception queue is prepared for each core. By constantly polling the transmission / reception queue assigned to each core, high-speed packet processing is realized. In accordance with this packet processing mechanism, it is necessary to implement timeout detection processing and reset processing.
Since the packet transmission queue is managed by the data plane core, the processing to detect that the status of the transmission queue does not change and determine that the situation has timed out for a predetermined time is implemented on the data plane core. There is a need. On the other hand, it is necessary to mount a process for performing recovery by instructing reset to the interface of the LAN card on the control plane core. For this reason, in order to realize these functions with the data plane development kit, cooperation between multiple cores is required. That is, the range of influence of the abnormality detection process and the restart process is widened, and there is a possibility that exclusive control and the like become complicated.

FIG. 3 is a diagram illustrating functional blocks of the network device 10.
As illustrated in FIG. 3, the network device 10 includes a control plane core processing unit 300 and a data plane core processing unit 400 as functional blocks of the LAN driver. Here, the control plane core and the control plane core processing unit are substantially the same. However, when attention is paid to hardware, the control plane core is referred to as a “control plane core”, and when attention is paid to functions or software, the control plane core is referred to as “control plane core”. It is called “core processing unit”. The relationship between the data plane core and the data plane core processing unit is the same.
The control plane core processing unit 300 performs control processing for the LAN card 120 and the like. The control plane core processing unit 300 of this example includes an upper engine processing unit 320, a LAN driver common unit 340, and a LAN driver 360. The upper engine processing unit 320 outputs data necessary for control to the LAN card 120 via the LAN driver common unit 340 and the LAN driver 360 in accordance with an instruction from the upper layer. The control plane core processing unit 300 is an example of a return instruction unit according to the present invention. When an abnormality of the LAN card 120 is detected, the control plane core processing unit 300 of this example outputs control information for sequentially executing the termination process, initialization process, and start process of the LAN card 120, and the LAN card 120. Realize the restart operation.

The data plane core processing unit 400 performs packet processing. The data plane core processing unit 400 of this example includes an upper engine processing unit 420, a LAN driver common unit 440, and a LAN driver 460. The upper engine processing unit 420 executes processing such as a firewall.
The LAN driver common unit 440 performs a process common to the LAN card on packets to be transmitted and received in response to an instruction from the upper engine processing unit 420. The LAN driver common unit 440 is an example of a transmission / reception unit according to the present invention, and outputs at least control information for prohibiting transmission processing to the LAN driver 460 while the restarting operation of the LAN card is being performed. More specifically, the LAN driver common unit 440 calls an idle function that makes the transmission / reception process idle.

  The LAN driver 460 directly accesses the LAN card 120 and transmits / receives a packet instructed from the LAN driver common unit 440. That is, the received packet is received from the LAN card 120 by the LAN driver 460, passes through the LAN driver common unit 440, and a unique function such as a firewall function is applied by the upper engine processing unit 420. The transmission packet passes from the upper engine processing unit 420 through the LAN driver common unit 440 and is transferred from the LAN driver 460 to the LAN card 120. The LAN driver 460 is an example of a lower-level transmission / reception unit according to the present invention.

FIG. 4 is a diagram illustrating a more detailed functional configuration of the LAN driver 360 (control plane core processing unit) and the LAN driver 460 (data plane core processing unit) in FIG.
As illustrated in FIG. 4, the LAN driver 360 provided in the control plane core processing unit 300 includes an access control unit 362, a dummy access control unit 364, a monitoring thread unit 366, and a reset processing unit 368.
The access control unit 362 controls each component involved in packet processing in response to a control request from the host. The access control unit 362 in this example implements a control function prepared in the data plane development kit.
Even if the dummy access control unit 364 receives a control request from the host while the reset processing unit 368 performs the restart operation of the LAN card 120, the control plane core processing unit 300 performs processing corresponding to this request. To do nothing.

  The monitoring thread unit 366 periodically checks the plurality of data plane core processing units 400 to monitor whether the transmission timeout detection flag is set. The monitoring thread unit 366 determines that an abnormality of the LAN card 120 has occurred when the transmission timeout detection flag is set in any of the data plane core processing units 400. The monitoring thread unit 366 is an example of a detection unit according to the present invention.

  When the monitoring thread unit 366 detects an abnormality of the LAN card 120, the reset processing unit 368 instructs the LAN card 120 to perform a restart operation. The reset processing unit 368 of this example outputs control information for sequentially executing termination processing, initialization processing, and start processing of the LAN card 120, and realizes a restart operation of the LAN card 120.

The LAN driver 460 provided in the data plane core processing unit 400 includes a transmission processing unit 462, a dummy transmission processing unit 466, a reception processing unit 468, and a dummy reception processing unit 470. The transmission processing unit 462 includes a timeout check unit 464.
The transmission processing unit 462 executes packet transmission processing in accordance with the data plane development kit. The transmission processing unit 462 of this example constantly polls the transmission queue assigned to each data plane core processing unit 400 and repeatedly executes packet transmission processing.
The timeout check unit 464 monitors the state of the transmission queue, and sets the transmission timeout detection flag when the state of the transmission queue for the predetermined time does not change.

  The dummy transmission processing unit 466 performs packet transmission processing idle according to the transmission idle function called by the LAN driver common unit 440. That is, the dummy transmission processing unit 466 returns to the upper layer that the packet could not be transmitted without accessing the LAN card 120 according to the transmission idle function called by the LAN driver common unit 440. As a result, the upper layer can detect that transmission is not possible, and can perform retry processing or packet drop handling. In other words, the hardware is not accessed while continuing the link-up state, and it is emulated as if it is being transmitted in a pseudo manner.

The reception processing unit 468 executes packet reception processing in accordance with the data plane development kit. Following the transmission processing unit 462, the reception processing unit 468 of this example repeatedly executes packet reception processing.
The dummy reception processing unit 470 performs packet reception processing idle according to the reception idle function called by the LAN driver common unit 440. That is, the dummy reception processing unit 470 returns to the upper layer that the packet could not be received without accessing the LAN card 120 in accordance with the reception idle function called by the LAN driver common unit 440. As a result, the upper layer can continue processing because there is no received packet.

FIG. 5A is a diagram illustrating a system configuration centering on the relationship between the control plane core and data plane core and the interface of the LAN card.
FIG. 5B is a diagram illustrating a system configuration when attention is paid to a single interface.
As illustrated in FIG. 5, one of the plurality of cores (# 0 to #N) of the CPU 100 is assigned as a control plane core, and the remaining N cores are assigned as data plane cores. Each data plane core executes packet transmission / reception processing by loop processing.
That is, in each data plane core, a transmission / reception queue is prepared for each interface of the LAN card 120, and high-speed packet processing is realized by executing transmission / reception processing in an infinite loop. In this high-speed packet processing, transmission processing and reception processing are executed in a loop. Of course, there is a case where there is no packet depending on timing, but such a case may exist as a “normal state”. Therefore, many configurations of packet transmission / reception processing can be continued in the “normal state”.
Thereby, even during the restart process of the LAN card 120, the influence range can be limited to the minimum necessary.

FIG. 6 is a diagram schematically illustrating packet transmission / reception processing by the data plane core processing unit 400 in a normal state.
FIG. 7 is a diagram schematically illustrating the state of the data plane core processing unit 400 during the restart of the LAN card.
As shown in FIG. 6, in each data plane core processing unit 400, a single thread linked to each core starts a packet transmission / reception function in an infinite loop. An instruction from the upper engine processing unit 420 is input to the lowermost LAN driver 460 via the LAN driver common unit 440 (LAN driver common unit in the figure), and the LAN driver 460 directly accesses the LAN card 120 to transmit the packet. Sending and receiving are performed continuously.
When the abnormality of the LAN card 120 is detected and the restart operation is started, as shown in FIG. 7, the LAN driver common unit 440 does not cause the LAN driver 460 to perform packet transmission / reception (transmission / reception idle function). ). As a result, the lower LAN driver 460 returns the result to the LAN driver common unit 440 without accessing the LAN card 120 being reset, and returns the result to the upper engine processing unit 420 as a result. In the upper engine processing unit 420, the same processing as that in the normal state (in the case of FIG. 6) is continued.
When the restart of the LAN card 120 is completed, the LAN driver common unit 440 returns the transmission / reception idle function to the normal function of FIG. 6 to return to the normal state.

FIG. 8 is a flowchart for explaining the abnormality detection process (S10) on the data plane core side.
As illustrated in FIG. 8, the time-out check unit 464 (FIG. 4) of the data plane core processing unit 400 determines whether there is a usable descriptor. If there is no usable descriptor (S100: Yes), the process of S110 is performed. If there is a usable descriptor (S100: No), the process proceeds to S105.

  In step 105 (S105), the timeout check unit 464 sets All F to the variable last_tx_end and sets false to the transmission timeout detection flag detect_tx_hung.

  In step 110 (S110), the timeout check unit 464 determines whether the variable tx_end does not match the variable last_tx_end. If the variable tx_end does not match the variable last_tx_end (S110: Yes), it is determined that the stop has been canceled and a different stop has occurred because the stop has been stopped at an index different from the transmission stop detected previously. If the variable tx_end matches the variable last_tx_end (S110: No), the timeout check unit 464 determines that the detected transmission stop is continuing, and proceeds to determination of the transmission stop duration (S120).

In step 115 (S115), the timeout check unit 464 sets the last entry index of the transmission queue to the variable last_tx_end, and sets the current time to the time tx_busy_time when the transmission descriptor acquisition failure is detected.
The processing returns to time S100.

  In step 120 (S120), the timeout check unit 464 determines whether or not the time during which the transmission queue does not change exceeds 2.9 seconds (S120: Yes). If the process proceeds to 2.9 seconds or less (S120: No), the process returns to S100.

In step 125 (S125), the timeout check unit 464 sets the transmission timeout detection flag detect_tx_hung to true and sets a flag.
As described above, the timeout check unit 464 of this example sets transmission timeout detection when the transmission queue does not change for 2.9 seconds after detecting that the transmission descriptor cannot be acquired. This is because the threshold for duplex switching has a value that transmission for 3 seconds is not possible, and it is assumed that the total processing time of the control plane core and data plane core must be within 3 seconds. This is because is made as long as possible.

FIG. 9 is a flowchart for explaining the abnormality detection process (S20) on the control plane core side.
As illustrated in FIG. 9, in step 200 (S200), the control plane core processing unit 300 initializes the transmission process together with the transmission timeout detection flag.
In step 205 (S205), the control plane core processing unit 300 periodically activates the monitoring thread unit 366 (FIG. 4).
In step 210 (S210), the activated monitoring thread unit 366 checks the management area of all the data plane core processing units 400.

  In step 215 (S215), the monitoring thread unit 366 determines whether the transmission timeout detection flag detect_tx_hung is set to true in any of the data plane core processing units 400, and any of the data plane core processing units 400. If the transmission timeout detection flag detect_tx_hung is set to true (S215: Yes), the process proceeds to S220. If any transmission timeout detection flag detect_tx_hung is set to false (S215: No), the process proceeds to S230. Transition to processing.

In step 220 (S220), the control plane core processing unit 300 determines whether or not the PAUSE frame is being received. If PAUSE frame is being received (S220: Yes), the control plane core processing unit 300 proceeds to the process of S230. Otherwise (S220: No), the process proceeds to S225.
In step 225 (S225), the monitoring thread unit 366 determines that an abnormality of the LAN card 120 has been detected, and instructs the reset processing unit 368 to perform reset processing.
The reset processing unit 368 instructs the LAN card 120 to restart. The control plane core processing unit 300 initializes resources for all data plane core processing units, and initializes variables and a transmission timeout flag to false.

In step 230 (S230), the monitoring thread unit 366 waits for 0.1 second and returns to the process of S210.
Thus, the control plane core processing unit 300 activates the monitoring thread unit 366 in order to check the transmission timeout detection flag of all the data plane core processing units 400. After startup, the monitoring thread unit 366 periodically checks the transmission timeout detection flag of all the data plane core processing units 400 at 0.1 second intervals. By doing so, the reset recovery can be executed within 3 seconds together with the 2.9 seconds required by the data plane core processing unit 400.

FIG. 10 is a flowchart for explaining the LAN card restart process (S30).
As illustrated in FIG. 10, in step 300 (S300), the control plane core processing unit 300 sets a reset flag.
In step 305 (S305), the control plane core processing unit 300 instructs the LAN driver common unit 440 of all the data plane core processing units 400 to set a transmission / reception idle function, and each data plane core processing unit 400. The LAN driver common unit 440 sets a transmission / reception idle function in the LAN driver 460. Thereby, the transmission processing unit 462 and the reception processing unit 468 stop the transmission / reception process, and the dummy transmission processing unit 466 and the dummy reception processing unit 470 start the transmission / reception idle process. As a result, the data plane core processing unit 400 does not access the LAN card 120.

In step 310 (S310), the dummy access control unit 364 of the control plane core processing unit 300 is set so as to make an unsupported response to a request from an upper layer, and is not supported in response to a request from another thread. It responds that it is.
In step 315 (S315), the control plane core processing unit 300 waits for 20 milliseconds and completes the processing in the current data plane processing unit 400.
In step 320 (S320), the control plane core processing unit 300 executes a close process. Specifically, the reset processing unit 368 causes the LAN card 120 to end.

In step 325 (S325), the control plane core processing unit 300 waits for 900 milliseconds. Since at least 700 milliseconds are required between link down and up, the waiting period is provided.
In step 330 (S330), the reset processing unit 368 of the control plane core processing unit 300 starts the initialization process of the LAN card 120. During the initialization process, the data plane core processing unit 400 is not allowed to set a transmission / reception function.
In step 335 (S335), the reset processing unit 368 executes configure processing for the LAN card 120.
In step 340 (S340), the reset processing unit 368 performs queue setup processing for the transmission / reception queue used for the transmission / reception processing on the LAN card 120.

In step 345 (S345), the reset processing unit 368 starts the start processing of the LAN card 120. Note that the transmission / reception function is not set in the data plane core processing unit 400 even during the start process.
In step 350 (S350), the reset processing unit 368 sets the promiscuous mode in the LAN card 120.

In step 355 (S355), the dummy access control unit 364 ends the unsupported response, and the access control unit 364 resumes the normal operation mode.
In step 360 (S360), the control plane core processing unit 300 instructs the LAN driver common unit 440 of all the data plane core processing units 400 to restore the transmission / reception function.
The LAN driver common unit 440 of each data plane core processing unit 400 returns the transmission / reception idle function to the normal transmission / reception function. That is, the dummy transmission processing unit 466 and the dummy reception processing unit 470 end the transmission / reception idle processing, and the transmission processing unit 462 and the reception processing unit 468 resume transmission / reception processing.

  In step 365 (S365), the control plane core processing unit 300 releases the reset flag and completes the restart processing of the LAN card 120.

Next, a specific example of the transmission / reception idle function will be described using an example of the system.
In the packet transmission / reception process of the data plane core processing unit 400, the LAN driver common unit 440 calls the function pointer (rx_pkt_burst, tx_pkt_burst) set in the global area I / F device structure as the transmission / reception function of the LAN driver 460. It is started with.
struct rte_eth_dev {
eth_rx_burst_t rx_pkt_burst; / ** <Pointer to PMD receive function. * /
eth_tx_burst_t tx_pkt_burst; / ** <Pointer to PMD transmit function. * /
struct rte_eth_dev_data * data; / ** <Pointer to device data * /
const struct eth_driver * driver; / ** <Driver for this device * /
struct eth_dev_ops * dev_ops; / ** <Functions exported by PMD * /
struct rte_pci_device * pci_dev; / ** <PCI info. supplied by probing * /
struct rte_eth_dev_cb_list callbacks; / ** <User application callbacks * /
};

The interface of the transmission / reception function is determined as follows, and passes necessary information such as a packet information area as an argument, and returns the number of received packets and the number of transmitted packets as return values.
typedef uint16_t (* eth_rx_burst_t) (void * rxq,
struct rte_mbuf ** rx_pkts,
uint16_t nb_pkts);
/ ** <@internal Retrieve input packets from a receive queue of an Ethernet device.
typedef uint16_t (* eth_tx_burst_t) (void * txq,
struct rte_mbuf ** tx_pkts,
uint16_t nb_pkts);
/ ** <@internal Send output packets on a transmit queue of an Ethernet device. * /

  This time, according to the above function interface, the following swing function was implemented. The reception function is activated and reception processing is performed, but it is shown to the upper layer that the packet received from the LAN card 120 does not exist. In addition, although the transmission function is activated and transmission processing is performed, it is shown to the upper layer that there is no packet that can be transmitted to the LAN card 120.

(Receiving processing idle function)
uint16_t
eth_igb_recv_pkts_reinit (__ rte_unused void * rx_queue,
__rte_unused struct rte_mbuf ** rx_pkts,
__rte_unused uint16_t nb_pkts)
{
/ * now resetting * /
return 0;
}
(Empty function for transmission processing)
uint16_t
eth_igb_xmit_pkts_reinit (__ rte_unused void * tx_queue,
__rte_unused struct rte_mbuf ** tx_pkts,
__rte_unused uint16_t nb_pkts)
{
/ * now resetting * /
return 0;
}

  The above is an implementation example of the 1 Gbps interface of the LAN card 120. In the case of the 10 Gbps interface, the following functions are registered in the same manner.

uint16_t
ixgbe_recv_pkts_reinit (__ rte_unused void * rx_queue,
__rte_unused struct rte_mbuf ** rx_pkts,
__rte_unused uint16_t nb_pkts)
{
/ * now resetting * /
return 0;
}
uint16_t
ixgbe_xmit_pkts_reinit (__ rte_unused void * tx_queue,
__rte_unused struct rte_mbuf ** tx_pkts,
__rte_unused uint16_t nb_pkts)
{
/ * now resetting * /
return 0;
}

Similarly, also in the control plane core processing unit 300, various control functions for the LAN driver 360 are activated when the LAN driver common unit 340 calls a function pointer set in the I / F device structure in the global area. . Setting these to “NULL” creates a state in which various control functions are not supported. For example, when changing the interface setting, a control function is registered in a function pointer called dev_configure. The following macro is used to start the control function. However, when the function pointer becomes “NULL”, -ENOTSUP is returned.
FUNC_PTR_OR_ERR_RET (* dev->dev_ops-> dev_configure, -ENOTSUP);

In the flowchart of FIG. 10, for example, various control functions are set to “NULL” in the “set to unsupport response to a request from another thread of the control plane core processing unit 300”, and “control plane Return to the normal control function at the point of “cancel unsupported response to request from other thread”.
Incidentally, in this embodiment, the control plane core processing unit 300 generates the number of monitoring thread units 366 for the number of interfaces, and there is a concern that these may affect the performance. It has been confirmed that the performance impact is negligible (0.7%) in the settings.

  As described above, when an abnormality occurs in the LAN card 120, the network device 10 according to the present embodiment can detect the abnormality and execute the restart operation of the LAN card 120 while limiting the range of influence. it can.

DESCRIPTION OF SYMBOLS 1 ... Network communication system 10 ... Network apparatus 100 ... CPU
DESCRIPTION OF SYMBOLS 102 ... Memory 104 ... Slot 120 ... LAN card 300 ... Control plane core process part 320 ... High-order engine process part 340 ... LAN driver common part 360 ... LAN driver 362 ... Access control part 364 ... Dummy access control part 366 ... Monitoring thread part 368 ... reset processing section 400 ... data plane core processing section 420 ... upper engine processing section 440 ... LAN driver common section 460 ... LAN driver 462 ... transmission processing section 464 ... timeout check section 466 ... dummy transmission processing section 468 ... reception processing section 470 ... Dummy reception processor

Claims (7)

A transmission / reception unit for transmitting / receiving data via a LAN card;
In accordance with control information from the transmission / reception unit, a lower-level transmission / reception unit that accesses the LAN card and transmits / receives data;
A detection unit for detecting an abnormality of the LAN card;
A recovery instruction unit for instructing a restart operation to the LAN card when an abnormality is detected by the detection unit;
The transmission / reception unit outputs at least control information for prohibiting transmission processing to the lower-level transmission / reception unit while the restart operation instructed by the return instruction unit is being performed. The information processing apparatus according to claim 1, wherein the detection unit detects an abnormality of the LAN card based on a state of a transmission queue managed by the transmission / reception unit. The return instruction unit is a single control plane core,
The transceiver unit includes a plurality of data plane cores,
The information processing apparatus according to claim 2, wherein all the data plane cores output control information that prohibits at least transmission processing to the lower transmission unit. The information processing apparatus according to claim 3, wherein the detection unit determines that the LAN card is abnormal when a transmission queue does not change for a predetermined period in any of the data plane cores. While the restart operation of the LAN card is being performed, the control plane core ignores the request from the upper layer, and sequentially performs the LAN card termination processing, initialization processing, and start processing to the lower transmission unit. The information processing apparatus according to claim 4, wherein control information to be executed is output to realize a restart operation of the LAN card. A transmission / reception step for performing data transmission / reception processing via a LAN card;
In accordance with control information from the transmission / reception step, a lower transmission / reception step of accessing the LAN card and transmitting / receiving data;
A detection step of detecting an abnormality of the LAN card;
A return instruction step for instructing a restart operation to the LAN card when an abnormality of the LAN card is detected;
An information processing method comprising: an output step of outputting at least control information for prohibiting transmission processing to a lower transmission / reception step of performing data transmission / reception processing while the restarting operation of the LAN card is being performed. A transmission / reception function for performing data transmission / reception processing via a LAN card;
In accordance with control information from the transmission / reception function, a lower transmission / reception function for accessing the LAN card and transmitting / receiving data;
A detection function for detecting an abnormality of the LAN card;
A return instruction function for instructing a restart operation to the LAN card when an abnormality of the LAN card is detected;
A program for causing a computer to realize at least an output function for outputting control information for prohibiting transmission processing to a lower transmission / reception function for performing data transmission / reception processing while the restart operation of the LAN card is being executed.

Images