Skip to contentSkip to conversation

mmc: rtsx: add support for pre_req and post_req

DISCUSSION Dan Carpenter 2 participants 3 messages Feb 24, 2014 – Feb 25, 2014 MMC, SD and SDIO

lore
Share & tools
lore
#0 lore raw

Hello Micky Ching,

This is a semi-automatic email about new static checker warnings.

The patch c42deffd5b53: "mmc: rtsx: add support for pre_req and post_req" from Feb 17, 2014, leads to the following Smatch complaint:

drivers/mmc/host/rtsx_pci_sdmmc.c:194 sd_finish_request()
error: we previously assumed 'mrq' could be null (see line 158)

drivers/mmc/host/rtsx_pci_sdmmc.c
157 mrq = host->mrq;

   158		if (!mrq) {
                    ^^^^

Patch introduces check.

159 dev_err(sdmmc_dev(host), "error: no request need finish\n");
160 goto out;
161 }
162
163 cmd = mrq->cmd;
164 data = mrq->data;
165
166 any_error = (mrq->sbc && mrq->sbc->error) ||
167 (mrq->stop && mrq->stop->error) ||
168 (cmd && cmd->error) || (data && data->error);
169
170 if (any_error) {
171 rtsx_pci_stop_cmd(pcr);
172 sd_clear_error(host);
173 }
174
175 if (data) {
176 if (any_error)
177 data->bytes_xfered = 0;
178 else
179 data->bytes_xfered = data->blocks * data->blksz;
180
181 if (!data->host_cookie)
182 rtsx_pci_dma_unmap_sg(pcr, data->sg, data->sg_len,
183 data->flags & MMC_DATA_READ);
184
185 }
186
187 host->mrq = NULL;
188 host->cmd = NULL;
189 host->data = NULL;
190
191 out:
192 spin_unlock_irqrestore(&host->lock, flags);
193 mutex_unlock(&pcr->pcr_mutex);

   194		mmc_request_done(host->mmc, mrq);
                                            ^^^

Dereferenced inside function. Patch introduces dereference.

195 }
196

regards,
dan carpenter

#1 lore raw

Hi Dan,

I'm wondering how to get this warning info, so I can check before sending patch. Can you give me some idea?

Best Regards.
micky.

Quoted 61 lines · Dan Carpenter

Hello Micky Ching,

This is a semi-automatic email about new static checker warnings.

The patch c42deffd5b53: "mmc: rtsx: add support for pre_req and post_req" from Feb 17, 2014, leads to the following Smatch complaint:

drivers/mmc/host/rtsx_pci_sdmmc.c:194 sd_finish_request()
error: we previously assumed 'mrq' could be null (see line 158)
drivers/mmc/host/rtsx_pci_sdmmc.c

    157		mrq = host->mrq;
    158		if (!mrq) {
                     ^^^^

Patch introduces check.

    159			dev_err(sdmmc_dev(host), "error: no request need finish\n");
    160			goto out;
    161		}
    162	
    163		cmd = mrq->cmd;
    164		data = mrq->data;
    165	
    166		any_error = (mrq->sbc && mrq->sbc->error) ||
    167			(mrq->stop && mrq->stop->error) ||
    168			(cmd && cmd->error) || (data && data->error);
    169	
    170		if (any_error) {
    171			rtsx_pci_stop_cmd(pcr);
    172			sd_clear_error(host);
    173		}
    174	
    175		if (data) {
    176			if (any_error)
    177				data->bytes_xfered = 0;
    178			else
    179				data->bytes_xfered = data->blocks * data->blksz;
    180	
    181			if (!data->host_cookie)
    182				rtsx_pci_dma_unmap_sg(pcr, data->sg, data->sg_len,
    183						data->flags & MMC_DATA_READ);
    184	
    185		}
    186	
    187		host->mrq = NULL;
    188		host->cmd = NULL;
    189		host->data = NULL;
    190	
    191	out:
    192		spin_unlock_irqrestore(&host->lock, flags);
    193		mutex_unlock(&pcr->pcr_mutex);
    194		mmc_request_done(host->mmc, mrq);
                                             ^^^

Dereferenced inside function. Patch introduces dereference.

    195	}
    196	

regards,
dan carpenter
.

#2 lore raw
Quoted 6 lines · micky

Hi Dan,

I'm wondering how to get this warning info, so I can check before sending patch.
Can you give me some idea?

This is a smatch thing, but you it's a cross function bug so you have to build the database first and that takes some hours. The commands are simple though. Run:

~/path/to/smatch_scripts/build_kernel_data.sh ~/path/to/smatch_scripts/kchecker drivers/mmc/host/rtsx_pci_sdmmc.c

regards,
dan carpenter