Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi all Just replied to remind this patch. Best Regards, Jaehoon Chung On 04/04/2013 03:49 AM, Simon Glass wrote: Hi Andy, On Sat, Mar 16, 2013 at 1:35 PM, Simon Glass s...@chromium.org wrote: Hi Andy, On Tue, Feb 12, 2013 at 7:14 PM, Jaehoon Chung jh80.ch...@samsung.com wrote: Hi Simon, It looks good to me. Acked-by: Jaehoon Chung jh80.ch...@samsung.com Do you think this will be picked up this time around? ping? Any thoughts on this one please? Regards, Simon ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi Andy, On Sat, Mar 16, 2013 at 1:35 PM, Simon Glass s...@chromium.org wrote: Hi Andy, On Tue, Feb 12, 2013 at 7:14 PM, Jaehoon Chung jh80.ch...@samsung.com wrote: Hi Simon, It looks good to me. Acked-by: Jaehoon Chung jh80.ch...@samsung.com Do you think this will be picked up this time around? ping? Any thoughts on this one please? Regards, Simon ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi Andy, On Tue, Feb 12, 2013 at 7:14 PM, Jaehoon Chung jh80.ch...@samsung.com wrote: Hi Simon, It looks good to me. Acked-by: Jaehoon Chung jh80.ch...@samsung.com Do you think this will be picked up this time around? Regards, Simon ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi Simon, It looks good to me. Acked-by: Jaehoon Chung jh80.ch...@samsung.com On 02/09/2013 01:48 AM, Simon Glass wrote: Hi, On Sun, Dec 16, 2012 at 7:44 PM, Simon Glass s...@chromium.org wrote: Hi, On Sun, Dec 16, 2012 at 6:12 PM, Jaehoon Chung jh80.ch...@samsung.com wrote: On 12/16/2012 02:18 AM, Simon Glass wrote: Hi, On Fri, Nov 30, 2012 at 3:13 PM, Simon Glass s...@chromium.org wrote: Hi Jaehoon, On Fri, Nov 30, 2012 at 12:25 AM, Jaehoon Chung jh80.ch...@samsung.com wrote: Hi, This concept is very good. But I have one question. I think need to call mmc_init() one more, right? how did you save the boot time(200ms)? On 11/29/2012 10:21 AM, Simon Glass wrote: From: Che-Liang Chiou clch...@chromium.org Most of time that MMC driver spends on initializing a device is polling OCR (operation conditions register). To decouple this polling loop, device init is split into two parts: The first part fires the OCR query command, and the second part polls the result. So the caller is now no longer bound to the OCR-polling delay; he may fire the query, go somewhere and then come back later for the result. To use this, call mmc_set_preinit() on any device which needs this. This can save significant amounts of time on boot (e.g. 200ms) by hiding the MMC init time behind other init. snip.. +int mmc_init(struct mmc *mmc) +{ + int err = IN_PROGRESS; + unsigned start = get_timer(0); + + if (mmc-has_init) + return 0; + if (!mmc-init_in_progress) + err = mmc_start_init(mmc); It need not to return? if err is IN_PROGRESS, next condition is immediately run. Then i think we didn't save the time before adjust this patch. It's a little confusing, but the way it works is that mmc_preinit() calls mmc_start_init() early in boot. Then when mmc_init() finally gets called (later) it finishes off the init. We still need mmc_init() to actually fully complete the init. If it were to return before completing the init then we would be unable to use the MMC. + + if (!err || err == IN_PROGRESS) + err = mmc_complete_init(mmc); + debug(%s: %d, time %lu\n, __func__, err, get_timer(start)); return err; } Does this patch look good now? I am wondering if it will be including in release, or in next? Well, concept is very good. But i didn't see the any benefit yet. I will test more...and share the result. OK thanks. Assuming that make sure that the pre-init is definitely done, I suggest you then make U-Boot go and do something else for 200ms, then do the full mmc init after that. It should complete immediately. So there is no time saving in mmc init, but you can do something else while waiting for the MMC init to complete (it takes 200ms or so for me). Another thing to note with eMMC is that we probe for SD first. This takes a little bit of time. We could perhaps use device tree to provide information about the type of attached memory, and thus avoid that probe. Does this patch look suitable for merging at this point? Regards, Simon ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi, On Sun, Dec 16, 2012 at 7:44 PM, Simon Glass s...@chromium.org wrote: Hi, On Sun, Dec 16, 2012 at 6:12 PM, Jaehoon Chung jh80.ch...@samsung.com wrote: On 12/16/2012 02:18 AM, Simon Glass wrote: Hi, On Fri, Nov 30, 2012 at 3:13 PM, Simon Glass s...@chromium.org wrote: Hi Jaehoon, On Fri, Nov 30, 2012 at 12:25 AM, Jaehoon Chung jh80.ch...@samsung.com wrote: Hi, This concept is very good. But I have one question. I think need to call mmc_init() one more, right? how did you save the boot time(200ms)? On 11/29/2012 10:21 AM, Simon Glass wrote: From: Che-Liang Chiou clch...@chromium.org Most of time that MMC driver spends on initializing a device is polling OCR (operation conditions register). To decouple this polling loop, device init is split into two parts: The first part fires the OCR query command, and the second part polls the result. So the caller is now no longer bound to the OCR-polling delay; he may fire the query, go somewhere and then come back later for the result. To use this, call mmc_set_preinit() on any device which needs this. This can save significant amounts of time on boot (e.g. 200ms) by hiding the MMC init time behind other init. snip.. +int mmc_init(struct mmc *mmc) +{ + int err = IN_PROGRESS; + unsigned start = get_timer(0); + + if (mmc-has_init) + return 0; + if (!mmc-init_in_progress) + err = mmc_start_init(mmc); It need not to return? if err is IN_PROGRESS, next condition is immediately run. Then i think we didn't save the time before adjust this patch. It's a little confusing, but the way it works is that mmc_preinit() calls mmc_start_init() early in boot. Then when mmc_init() finally gets called (later) it finishes off the init. We still need mmc_init() to actually fully complete the init. If it were to return before completing the init then we would be unable to use the MMC. + + if (!err || err == IN_PROGRESS) + err = mmc_complete_init(mmc); + debug(%s: %d, time %lu\n, __func__, err, get_timer(start)); return err; } Does this patch look good now? I am wondering if it will be including in release, or in next? Well, concept is very good. But i didn't see the any benefit yet. I will test more...and share the result. OK thanks. Assuming that make sure that the pre-init is definitely done, I suggest you then make U-Boot go and do something else for 200ms, then do the full mmc init after that. It should complete immediately. So there is no time saving in mmc init, but you can do something else while waiting for the MMC init to complete (it takes 200ms or so for me). Another thing to note with eMMC is that we probe for SD first. This takes a little bit of time. We could perhaps use device tree to provide information about the type of attached memory, and thus avoid that probe. Does this patch look suitable for merging at this point? Regards, Simon ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
On 12/16/2012 02:18 AM, Simon Glass wrote: Hi, On Fri, Nov 30, 2012 at 3:13 PM, Simon Glass s...@chromium.org wrote: Hi Jaehoon, On Fri, Nov 30, 2012 at 12:25 AM, Jaehoon Chung jh80.ch...@samsung.com wrote: Hi, This concept is very good. But I have one question. I think need to call mmc_init() one more, right? how did you save the boot time(200ms)? On 11/29/2012 10:21 AM, Simon Glass wrote: From: Che-Liang Chiou clch...@chromium.org Most of time that MMC driver spends on initializing a device is polling OCR (operation conditions register). To decouple this polling loop, device init is split into two parts: The first part fires the OCR query command, and the second part polls the result. So the caller is now no longer bound to the OCR-polling delay; he may fire the query, go somewhere and then come back later for the result. To use this, call mmc_set_preinit() on any device which needs this. This can save significant amounts of time on boot (e.g. 200ms) by hiding the MMC init time behind other init. snip.. +int mmc_init(struct mmc *mmc) +{ + int err = IN_PROGRESS; + unsigned start = get_timer(0); + + if (mmc-has_init) + return 0; + if (!mmc-init_in_progress) + err = mmc_start_init(mmc); It need not to return? if err is IN_PROGRESS, next condition is immediately run. Then i think we didn't save the time before adjust this patch. It's a little confusing, but the way it works is that mmc_preinit() calls mmc_start_init() early in boot. Then when mmc_init() finally gets called (later) it finishes off the init. We still need mmc_init() to actually fully complete the init. If it were to return before completing the init then we would be unable to use the MMC. + + if (!err || err == IN_PROGRESS) + err = mmc_complete_init(mmc); + debug(%s: %d, time %lu\n, __func__, err, get_timer(start)); return err; } Does this patch look good now? I am wondering if it will be including in release, or in next? Well, concept is very good. But i didn't see the any benefit yet. I will test more...and share the result. Best Regards, Jaehoon Chung Regards, Simon [snip] Regards, Simon ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi, On Sun, Dec 16, 2012 at 6:12 PM, Jaehoon Chung jh80.ch...@samsung.com wrote: On 12/16/2012 02:18 AM, Simon Glass wrote: Hi, On Fri, Nov 30, 2012 at 3:13 PM, Simon Glass s...@chromium.org wrote: Hi Jaehoon, On Fri, Nov 30, 2012 at 12:25 AM, Jaehoon Chung jh80.ch...@samsung.com wrote: Hi, This concept is very good. But I have one question. I think need to call mmc_init() one more, right? how did you save the boot time(200ms)? On 11/29/2012 10:21 AM, Simon Glass wrote: From: Che-Liang Chiou clch...@chromium.org Most of time that MMC driver spends on initializing a device is polling OCR (operation conditions register). To decouple this polling loop, device init is split into two parts: The first part fires the OCR query command, and the second part polls the result. So the caller is now no longer bound to the OCR-polling delay; he may fire the query, go somewhere and then come back later for the result. To use this, call mmc_set_preinit() on any device which needs this. This can save significant amounts of time on boot (e.g. 200ms) by hiding the MMC init time behind other init. snip.. +int mmc_init(struct mmc *mmc) +{ + int err = IN_PROGRESS; + unsigned start = get_timer(0); + + if (mmc-has_init) + return 0; + if (!mmc-init_in_progress) + err = mmc_start_init(mmc); It need not to return? if err is IN_PROGRESS, next condition is immediately run. Then i think we didn't save the time before adjust this patch. It's a little confusing, but the way it works is that mmc_preinit() calls mmc_start_init() early in boot. Then when mmc_init() finally gets called (later) it finishes off the init. We still need mmc_init() to actually fully complete the init. If it were to return before completing the init then we would be unable to use the MMC. + + if (!err || err == IN_PROGRESS) + err = mmc_complete_init(mmc); + debug(%s: %d, time %lu\n, __func__, err, get_timer(start)); return err; } Does this patch look good now? I am wondering if it will be including in release, or in next? Well, concept is very good. But i didn't see the any benefit yet. I will test more...and share the result. OK thanks. Assuming that make sure that the pre-init is definitely done, I suggest you then make U-Boot go and do something else for 200ms, then do the full mmc init after that. It should complete immediately. So there is no time saving in mmc init, but you can do something else while waiting for the MMC init to complete (it takes 200ms or so for me). Another thing to note with eMMC is that we probe for SD first. This takes a little bit of time. We could perhaps use device tree to provide information about the type of attached memory, and thus avoid that probe. Regards, Simon Best Regards, Jaehoon Chung Regards, Simon [snip] Regards, Simon ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi, On Fri, Nov 30, 2012 at 3:13 PM, Simon Glass s...@chromium.org wrote: Hi Jaehoon, On Fri, Nov 30, 2012 at 12:25 AM, Jaehoon Chung jh80.ch...@samsung.com wrote: Hi, This concept is very good. But I have one question. I think need to call mmc_init() one more, right? how did you save the boot time(200ms)? On 11/29/2012 10:21 AM, Simon Glass wrote: From: Che-Liang Chiou clch...@chromium.org Most of time that MMC driver spends on initializing a device is polling OCR (operation conditions register). To decouple this polling loop, device init is split into two parts: The first part fires the OCR query command, and the second part polls the result. So the caller is now no longer bound to the OCR-polling delay; he may fire the query, go somewhere and then come back later for the result. To use this, call mmc_set_preinit() on any device which needs this. This can save significant amounts of time on boot (e.g. 200ms) by hiding the MMC init time behind other init. snip.. +int mmc_init(struct mmc *mmc) +{ + int err = IN_PROGRESS; + unsigned start = get_timer(0); + + if (mmc-has_init) + return 0; + if (!mmc-init_in_progress) + err = mmc_start_init(mmc); It need not to return? if err is IN_PROGRESS, next condition is immediately run. Then i think we didn't save the time before adjust this patch. It's a little confusing, but the way it works is that mmc_preinit() calls mmc_start_init() early in boot. Then when mmc_init() finally gets called (later) it finishes off the init. We still need mmc_init() to actually fully complete the init. If it were to return before completing the init then we would be unable to use the MMC. + + if (!err || err == IN_PROGRESS) + err = mmc_complete_init(mmc); + debug(%s: %d, time %lu\n, __func__, err, get_timer(start)); return err; } Does this patch look good now? I am wondering if it will be including in release, or in next? Regards, Simon [snip] Regards, Simon ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi, This concept is very good. But I have one question. I think need to call mmc_init() one more, right? how did you save the boot time(200ms)? On 11/29/2012 10:21 AM, Simon Glass wrote: From: Che-Liang Chiou clch...@chromium.org Most of time that MMC driver spends on initializing a device is polling OCR (operation conditions register). To decouple this polling loop, device init is split into two parts: The first part fires the OCR query command, and the second part polls the result. So the caller is now no longer bound to the OCR-polling delay; he may fire the query, go somewhere and then come back later for the result. To use this, call mmc_set_preinit() on any device which needs this. This can save significant amounts of time on boot (e.g. 200ms) by hiding the MMC init time behind other init. snip.. +int mmc_init(struct mmc *mmc) +{ + int err = IN_PROGRESS; + unsigned start = get_timer(0); + + if (mmc-has_init) + return 0; + if (!mmc-init_in_progress) + err = mmc_start_init(mmc); It need not to return? if err is IN_PROGRESS, next condition is immediately run. Then i think we didn't save the time before adjust this patch. + + if (!err || err == IN_PROGRESS) + err = mmc_complete_init(mmc); + debug(%s: %d, time %lu\n, __func__, err, get_timer(start)); return err; } @@ -1315,6 +1368,25 @@ int get_mmc_num(void) return cur_dev_num; } +void mmc_set_preinit(struct mmc *mmc, int preinit) +{ + mmc-preinit = preinit; +} + +static void do_preinit(void) +{ + struct mmc *m; + struct list_head *entry; + + list_for_each(entry, mmc_devices) { + m = list_entry(entry, struct mmc, link); + + if (m-preinit) + mmc_start_init(m); + } +} + + int mmc_initialize(bd_t *bis) { INIT_LIST_HEAD (mmc_devices); @@ -1325,5 +1397,6 @@ int mmc_initialize(bd_t *bis) print_mmc_devices(','); + do_preinit(); return 0; } diff --git a/include/mmc.h b/include/mmc.h index a13e2bd..445d714 100644 --- a/include/mmc.h +++ b/include/mmc.h @@ -62,6 +62,7 @@ #define UNUSABLE_ERR -17 /* Unusable Card */ #define COMM_ERR -18 /* Communications Error */ #define TIMEOUT -19 +#define IN_PROGRESS -20 /* operation is in progress */ #define MMC_CMD_GO_IDLE_STATE0 #define MMC_CMD_SEND_OP_COND 1 @@ -260,6 +261,10 @@ struct mmc { int (*init)(struct mmc *mmc); int (*getcd)(struct mmc *mmc); uint b_max; + char op_cond_pending; /* 1 if we are waiting on an op_cond command */ + char init_in_progress; /* 1 if we have done mmc_start_init() */ + char preinit; /* start init as early as possible */ + uint op_cond_response; /* the response byte from the last op_cond */ }; int mmc_register(struct mmc *mmc); @@ -276,6 +281,31 @@ int mmc_switch_part(int dev_num, unsigned int part_num); int mmc_getcd(struct mmc *mmc); void spl_mmc_load(void) __noreturn; +/** + * Start device initialization and return immediately; it does not block on + * polling OCR (operation condition register) status. Then you should call + * mmc_init, which would block on polling OCR status and complete the device + * initializatin. + * + * @param mmcPointer to a MMC device struct + * @return 0 on success, IN_PROGRESS on waiting for OCR status, 0 on error. + */ +int mmc_start_init(struct mmc *mmc); + +/** + * Set preinit flag of mmc device. + * + * This will cause the device to be pre-inited during mmc_initialize(), + * which may save boot time if the device is not accessed until later. + * Some eMMC devices take 200-300ms to init, but unfortunately they + * must be sent a series of commands to even get them to start preparing + * for operation. + * + * @param mmcPointer to a MMC device struct + * @param preinitpreinit flag value + */ +void mmc_set_preinit(struct mmc *mmc, int preinit); + #ifdef CONFIG_GENERIC_MMC #define mmc_host_is_spi(mmc) ((mmc)-host_caps MMC_MODE_SPI) struct mmc *mmc_spi_init(uint bus, uint cs, uint speed, uint mode); ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi, On Thu, Nov 29, 2012 at 7:20 AM, Jae hoon Chung jh80.ch...@gmail.com wrote: Hi Simon, Is it saved the 200ms? Could you tell me your environment? I will check this patch..and share the result. The environment is snow (Samsung ARM Chromebook). The time save only comes from not waiting for the MMC init. So: 1. Kick off MMC init 2. Go off and do something else for 200ms 3. Come back and MMC init should complete immediately Regards, Simon Best Regards, Jaehoon Chung 2012/11/29 Simon Glass s...@chromium.org: From: Che-Liang Chiou clch...@chromium.org Most of time that MMC driver spends on initializing a device is polling OCR (operation conditions register). To decouple this polling loop, device init is split into two parts: The first part fires the OCR query command, and the second part polls the result. So the caller is now no longer bound to the OCR-polling delay; he may fire the query, go somewhere and then come back later for the result. To use this, call mmc_set_preinit() on any device which needs this. This can save significant amounts of time on boot (e.g. 200ms) by hiding the MMC init time behind other init. Signed-off-by: Che-Liang Chiou clch...@chromium.org Signed-off-by: Simon Glass s...@chromium.org --- Changes in v2: - Rebase to master drivers/mmc/mmc.c | 137 include/mmc.h | 30 2 files changed, 135 insertions(+), 32 deletions(-) diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c index 72e8ce6..09695e8 100644 --- a/drivers/mmc/mmc.c +++ b/drivers/mmc/mmc.c @@ -503,48 +503,70 @@ static int sd_send_op_cond(struct mmc *mmc) return 0; } -static int mmc_send_op_cond(struct mmc *mmc) +/* We pass in the cmd since otherwise the init seems to fail */ +static int mmc_send_op_cond_iter(struct mmc *mmc, struct mmc_cmd *cmd, + int use_arg) { - int timeout = 1; - struct mmc_cmd cmd; int err; + cmd-cmdidx = MMC_CMD_SEND_OP_COND; + cmd-resp_type = MMC_RSP_R3; + cmd-cmdarg = 0; + if (use_arg !mmc_host_is_spi(mmc)) { + cmd-cmdarg = + (mmc-voltages + (mmc-op_cond_response OCR_VOLTAGE_MASK)) | + (mmc-op_cond_response OCR_ACCESS_MODE); + + if (mmc-host_caps MMC_MODE_HC) + cmd-cmdarg |= OCR_HCS; + } + err = mmc_send_cmd(mmc, cmd, NULL); + if (err) + return err; + mmc-op_cond_response = cmd-response[0]; + return 0; +} + +int mmc_send_op_cond(struct mmc *mmc) +{ + struct mmc_cmd cmd; + int err, i; + /* Some cards seem to need this */ mmc_go_idle(mmc); /* Asking to the card its capabilities */ - cmd.cmdidx = MMC_CMD_SEND_OP_COND; - cmd.resp_type = MMC_RSP_R3; - cmd.cmdarg = 0; - - err = mmc_send_cmd(mmc, cmd, NULL); + mmc-op_cond_pending = 1; + for (i = 0; i 2; i++) { + err = mmc_send_op_cond_iter(mmc, cmd, i != 0); + if (err) + return err; - if (err) - return err; + /* exit if not busy (flag seems to be inverted) */ + if (mmc-op_cond_response OCR_BUSY) + return 0; + } + return IN_PROGRESS; +} - udelay(1000); +int mmc_complete_op_cond(struct mmc *mmc) +{ + struct mmc_cmd cmd; + int timeout = 1000; + uint start; + int err; + mmc-op_cond_pending = 0; + start = get_timer(0); do { - cmd.cmdidx = MMC_CMD_SEND_OP_COND; - cmd.resp_type = MMC_RSP_R3; - cmd.cmdarg = (mmc_host_is_spi(mmc) ? 0 : - (mmc-voltages - (cmd.response[0] OCR_VOLTAGE_MASK)) | - (cmd.response[0] OCR_ACCESS_MODE)); - - if (mmc-host_caps MMC_MODE_HC) - cmd.cmdarg |= OCR_HCS; - - err = mmc_send_cmd(mmc, cmd, NULL); - + err = mmc_send_op_cond_iter(mmc, cmd, 1); if (err) return err; - - udelay(1000); - } while (!(cmd.response[0] OCR_BUSY) timeout--); - - if (timeout = 0) - return UNUSABLE_ERR; + if (get_timer(start) timeout) + return UNUSABLE_ERR; + udelay(100); + } while (!(mmc-op_cond_response OCR_BUSY)); if (mmc_host_is_spi(mmc)) { /* read OCR for spi */ cmd.cmdidx = MMC_CMD_SPI_READ_OCR; @@ -1227,7 +1249,7 @@ block_dev_desc_t *mmc_get_dev(int dev) } #endif -int mmc_init(struct mmc *mmc) +int mmc_start_init(struct mmc *mmc) { int err; @@ -1267,17 +1289,48 @@ int mmc_init(struct
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi Jaehoon, On Fri, Nov 30, 2012 at 12:25 AM, Jaehoon Chung jh80.ch...@samsung.com wrote: Hi, This concept is very good. But I have one question. I think need to call mmc_init() one more, right? how did you save the boot time(200ms)? On 11/29/2012 10:21 AM, Simon Glass wrote: From: Che-Liang Chiou clch...@chromium.org Most of time that MMC driver spends on initializing a device is polling OCR (operation conditions register). To decouple this polling loop, device init is split into two parts: The first part fires the OCR query command, and the second part polls the result. So the caller is now no longer bound to the OCR-polling delay; he may fire the query, go somewhere and then come back later for the result. To use this, call mmc_set_preinit() on any device which needs this. This can save significant amounts of time on boot (e.g. 200ms) by hiding the MMC init time behind other init. snip.. +int mmc_init(struct mmc *mmc) +{ + int err = IN_PROGRESS; + unsigned start = get_timer(0); + + if (mmc-has_init) + return 0; + if (!mmc-init_in_progress) + err = mmc_start_init(mmc); It need not to return? if err is IN_PROGRESS, next condition is immediately run. Then i think we didn't save the time before adjust this patch. It's a little confusing, but the way it works is that mmc_preinit() calls mmc_start_init() early in boot. Then when mmc_init() finally gets called (later) it finishes off the init. We still need mmc_init() to actually fully complete the init. If it were to return before completing the init then we would be unable to use the MMC. + + if (!err || err == IN_PROGRESS) + err = mmc_complete_init(mmc); + debug(%s: %d, time %lu\n, __func__, err, get_timer(start)); return err; } [snip] Regards, Simon ___ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
Re: [U-Boot] [PATCH v2] mmc: Split device init to decouple OCR-polling delay
Hi Simon, Is it saved the 200ms? Could you tell me your environment? I will check this patch..and share the result. Best Regards, Jaehoon Chung 2012/11/29 Simon Glass s...@chromium.org: From: Che-Liang Chiou clch...@chromium.org Most of time that MMC driver spends on initializing a device is polling OCR (operation conditions register). To decouple this polling loop, device init is split into two parts: The first part fires the OCR query command, and the second part polls the result. So the caller is now no longer bound to the OCR-polling delay; he may fire the query, go somewhere and then come back later for the result. To use this, call mmc_set_preinit() on any device which needs this. This can save significant amounts of time on boot (e.g. 200ms) by hiding the MMC init time behind other init. Signed-off-by: Che-Liang Chiou clch...@chromium.org Signed-off-by: Simon Glass s...@chromium.org --- Changes in v2: - Rebase to master drivers/mmc/mmc.c | 137 include/mmc.h | 30 2 files changed, 135 insertions(+), 32 deletions(-) diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c index 72e8ce6..09695e8 100644 --- a/drivers/mmc/mmc.c +++ b/drivers/mmc/mmc.c @@ -503,48 +503,70 @@ static int sd_send_op_cond(struct mmc *mmc) return 0; } -static int mmc_send_op_cond(struct mmc *mmc) +/* We pass in the cmd since otherwise the init seems to fail */ +static int mmc_send_op_cond_iter(struct mmc *mmc, struct mmc_cmd *cmd, + int use_arg) { - int timeout = 1; - struct mmc_cmd cmd; int err; + cmd-cmdidx = MMC_CMD_SEND_OP_COND; + cmd-resp_type = MMC_RSP_R3; + cmd-cmdarg = 0; + if (use_arg !mmc_host_is_spi(mmc)) { + cmd-cmdarg = + (mmc-voltages + (mmc-op_cond_response OCR_VOLTAGE_MASK)) | + (mmc-op_cond_response OCR_ACCESS_MODE); + + if (mmc-host_caps MMC_MODE_HC) + cmd-cmdarg |= OCR_HCS; + } + err = mmc_send_cmd(mmc, cmd, NULL); + if (err) + return err; + mmc-op_cond_response = cmd-response[0]; + return 0; +} + +int mmc_send_op_cond(struct mmc *mmc) +{ + struct mmc_cmd cmd; + int err, i; + /* Some cards seem to need this */ mmc_go_idle(mmc); /* Asking to the card its capabilities */ - cmd.cmdidx = MMC_CMD_SEND_OP_COND; - cmd.resp_type = MMC_RSP_R3; - cmd.cmdarg = 0; - - err = mmc_send_cmd(mmc, cmd, NULL); + mmc-op_cond_pending = 1; + for (i = 0; i 2; i++) { + err = mmc_send_op_cond_iter(mmc, cmd, i != 0); + if (err) + return err; - if (err) - return err; + /* exit if not busy (flag seems to be inverted) */ + if (mmc-op_cond_response OCR_BUSY) + return 0; + } + return IN_PROGRESS; +} - udelay(1000); +int mmc_complete_op_cond(struct mmc *mmc) +{ + struct mmc_cmd cmd; + int timeout = 1000; + uint start; + int err; + mmc-op_cond_pending = 0; + start = get_timer(0); do { - cmd.cmdidx = MMC_CMD_SEND_OP_COND; - cmd.resp_type = MMC_RSP_R3; - cmd.cmdarg = (mmc_host_is_spi(mmc) ? 0 : - (mmc-voltages - (cmd.response[0] OCR_VOLTAGE_MASK)) | - (cmd.response[0] OCR_ACCESS_MODE)); - - if (mmc-host_caps MMC_MODE_HC) - cmd.cmdarg |= OCR_HCS; - - err = mmc_send_cmd(mmc, cmd, NULL); - + err = mmc_send_op_cond_iter(mmc, cmd, 1); if (err) return err; - - udelay(1000); - } while (!(cmd.response[0] OCR_BUSY) timeout--); - - if (timeout = 0) - return UNUSABLE_ERR; + if (get_timer(start) timeout) + return UNUSABLE_ERR; + udelay(100); + } while (!(mmc-op_cond_response OCR_BUSY)); if (mmc_host_is_spi(mmc)) { /* read OCR for spi */ cmd.cmdidx = MMC_CMD_SPI_READ_OCR; @@ -1227,7 +1249,7 @@ block_dev_desc_t *mmc_get_dev(int dev) } #endif -int mmc_init(struct mmc *mmc) +int mmc_start_init(struct mmc *mmc) { int err; @@ -1267,17 +1289,48 @@ int mmc_init(struct mmc *mmc) if (err == TIMEOUT) { err = mmc_send_op_cond(mmc); - if (err) { + if (err err != IN_PROGRESS) { printf(Card did not respond to voltage select!\n); return UNUSABLE_ERR; } } - err =